Generating spatial audio from portable products – Part 2: Acoustic beamforming using the LM48901
[Part 1 of this article introduces spatial audio reproduction and defines head related transfer function (HRTF), crosstalk cancellation and audio beamforming.]
Acoustic Beamforming Techniques
There are generally two types of acoustic beamforming techniques used today:
- Mechanical beamformers, which rely on the physical sizes and positions of the speakers to produce desired spatial effects.
- Electronic beamformers which rely on signal processing that is performed by a digital signal processor (DSP) before audio signals are provided to the speakers.
Of course, both methods may be combined as needed depending upon the application.
Electronic beamforming was initially developed for radar applications. Its application to audio first appeared in microphone arrays designed for speech and audio capture. The abundant applications in this area have led to many years of innovations in audio beamforming algorithms.
The basic idea of microphone array beamforming is to individually adjust the phase and amplitude of the received signal of each array element so that the combined output can achieve maximum signal-to-noise ratio (SNR) in certain directions. The concept is similar to extracting the desired signal in the frequency domain by bandpass filtering, but here it’s done in the spatial domain and the passband can be considered as a range of directions. Many beamforming techniques exist and are well documented, and the selection of certain technique usually depends on the requirements and constrains of a specific application.
Although audio beamforming is widely used in capturing audio signals, its application in audio playback is relatively limited. A major reason is that stereo systems have been delivering relatively good performance and it was unnecessary to use more than two speakers for many applications.
However, the ever shrinking mechanical space in portable devices now poses major challenges for stereo audio playback. For example, output volume level loss due to speaker size reduction, and diminished stereo sound image due to narrow speaker spacing are both problematic.
Flat-panel TV’s are trending toward very thin enclosures that severely limit loudspeaker cone excursion, which in turn diminishes output volume levels as well as audio quality. One way to overcome these limitations is to use an array of small speakers to increase the overall volume and to render a more desirable sound field using audio beamforming techniques.
It should be pointed out that there are many other methods that can deliver spatial audio over a speaker array system, such as wave field synthesis (WFS) and ambisonics. They typically require dozens to hundreds of speakers and a large span in space. Therefore they are mostly used to design speaker systems in theaters and sound rooms, but are not suitable for small to mid-size speaker arrays.
Generally speaking, beamforming techniques for microphone array applications can be readily adapted to speaker array applications since the playback is basically a reverse of the capturing process. However, full bandwidth audio playback requires much wider bandwidth than typical speech applications and therefore requires special considerations in algorithm choice and array setup.
To deliver spatial stereo sound over a speaker array requires a very effective crosstalk cancellation algorithm. Additionally, speaker array algorithms need to minimize the distortion of audio playback as much as possible, including minimizing artifacts and coloration.
As a quick review, generating spatial audio effects utilizes three techniques as seen in Figure 4: crosstalk cancellation, mechanical and electronic beamforming, and HRTF data.
Figure 4. Spatial audio technology uses a combination of crosstalk cancellation, head related transfer function, and beamforming.
The challenge has been to develop techniques suitable for small arrays which will yield subjectively better results than using stereo alone, and apply those in a manner that use several speakers without requiring complex algorithmic programming skills.
Spatial Array Amplifier
Texas Instruments has addressed this challenge with the LM48901, the first in a series of spatial array audio amplifiers for implementing loudspeaker arrays to produce an immersive audio experience for space-constrained applications.
The LM48901 is a four-channel Class D audio IC implementing distributed, electronic beamforming algorithms, plus sound placement using HRTF data to create beamformed loudspeaker arrays. Audio input can be either stereo digital audio (I2S format) or analog stereo audio. The LM48901 array processing system is non-complementary, meaning it does not require encoded stereo source material to produce the loudspeaker array spatial effects.
Electronic beamforming functionality is distributed across the various array filters placed within or connected to the various audio amplifiers. The filtered audio signals are coupled to one or more loudspeaker amplifiers to implement beamforming and produce desired spatial effects. The LM48901’s basic block diagram is shown in Figure 5.
Figure 5. The LM48901 integrates a spatial processing DSP, four Class D amplifiers, 18-bit stereo analog-to-digital converter (ADC), phase-locked loop (PLL), and I2S and I2C interfaces.
When using the HRTFs in the pre-filter stage, the horizontal span of the soundstage can easily be controlled by loading HRTFs that correspond to different angles. The overall effect generated by the two stage processing creates a pair of virtual speakers in front of the listener with adjustable separation between the two speakers.
As a result, a less powerful centralized DSP (and in some cases, no centralized DSP) is needed in an audio system driving loudspeaker arrays. The distributed processing also allows for more flexibility in the geometric arrangement of speaker arrays, as well as the number of speakers in the array.
The LM48901 provides four filterless Class D amplifiers capable of up to 1.7W per channel into 8-Ω speakers, or 2.8W per channel into 4-Ω speakers for a combined four-channel output power of 6.8W and 11.2W, respectively, using a 5V supply voltage. These output power levels are sufficient for many consumer audio applications. Lower output power levels are achieved with lower supply voltages down to a minimum of 2.7V. Overall efficiency exceeds 80% with 4-Ω loads, and 85% with 8-Ω loads.
Two of the channels can be connected in parallel to increase output power for applications requiring a subwoofer. In this way, a single LM48901 can implement a spatialized 2.1-channel audio system with three speakers.
In addition, several LM48901 devices can be daisy-chained together to support more loudspeakers, with a maximum of four loudspeakers supported per device. This feature greatly simplifies the design and implementation of speaker arrays of any size up to 64 loudspeakers.
The LM48901 filter coefficients can be updated dynamically in order to provide different beamforming functions. This can be useful when the configuration of a speaker array changes, such as when speakers are added to or removed from the array. The LM48901’s scalable speaker array technology easily supports multi-dimensional speaker array configurations.
Figure 6 shows a web-based design tool (https://webench.national.com/sawa/#main) that generates loudspeaker array filter coefficients based on user inputs about the array geometry and listener position. Use of the web design tool requires user registration.
Figure 6. TI’s web-based spatial audio software tool includes an easy-to-use speaker array coefficient generator that creates unique spatial audio coefficients in a few easy steps.
During the array design process, the user takes the following steps:
- Click on "New Config" to start a new configuration.
- If this design is a cross-over-design, set the cross-over-frequency value for the design. Non-cross-over-design has a minimum of 4 tweeter speakers. Cross-over design has a minimum of 4 tweeter speakers and 2 woofer speakers.
- Choose the sampling frequency for your design.
- Choose the number of speakers for your design.
- Set the Spatial effect Beamforming Bypass Frequency range. Note: Use a lower value to achieve more spatial effect; or a larger value to achieve more solid center sound.
- Specify the Listener Sweet Spot co-ordinate values.
- Apply the above settings to create Speaker Table listing speaker details and a co-ordinate system.
- Edit the speaker table entries or move the speaker images in Top View or Front View to create desired speaker array configuration.
- Press Generate Coefficients button to create the coefficients file for your designed configuration.
The web based design tool eliminates the need for the system engineer to program specific acoustic beamforming and/or HRTF positioning algorithms, or to use general purpose DSP programming language and system design kits.
A very useful feature of the design tool allows the user to click on speaker icons in the array, or the listener icon, and drag and drop them to the design geometric configuration. During the system design process the filter coefficients are then downloaded into the coefficient memory of the LM48901. The LM48901 based system is then tested for the desired results. The design can easily be changed by specifying different array parameters and reloading the coefficients.
A separate equalization tool can be downloaded that permits further frequency equalization across the speaker array for tuning the output audio quality. While the LM48901 simplifies creating the spatial effects using loudspeaker arrays, it does not eliminate the need to do careful acoustic design in the final product. Nor will it adequately compensate for using poor performance speakers in the array.
Spatial array technology video’s showing more information about the product and the web-based design tool can be found at https://www.ti.com/spatial-pr.
SUMMARY
Spatial audio techniques to improve clarity and depth are becoming more widespread in consumer audio applications spanning smartphones to large flat-screen TV’s. The LM48901 spatial array IC provides a very simple and fast way to implement those techniques in loudspeaker arrays to expand the soundstage and output power for products with limited space for speakers.
A single LM48901, with its accurate sound beamforming and crosstalk cancellation, can be used in two, three or four speaker applications. Daisy-chaining several LM48901 ICs enables the design of eight, 12, and 16 speaker arrays that deliver a much wider and exciting audio experience.
About the author:
Kenneth Boyce is Audio Technologist for Texas Instruments Silicon Valley Labs. He previously served as Technology Director for National Semiconductor’s Audio Products Group, and before joining National, Boyce served as director of the Audio and Communications Division at Oak Technology, which developed initial implementations of AC-97 Codecs and Digital Audio Controllers. He holds a bachelor of science degree in electronics from West Virginia University.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
