The combination of advanced, low-cost, high-performance silicon and constantly evolving software techniques has brought high-end audio to economy-level vehicles. However, the increasing integration of infotainment systems with high-end audio demands is putting audio signal processing designers under pressure to meet these demands while containing costs and keeping power consumption to a minimum.
For audiophiles, the Holy Grail is still to have as clean and unobstructed path as possible between the music playback source and the speakers to reproduce the musical soundstage as accurately as possible. This implies high-end audio components, encompassing the playback system’s audio pick-up, pre-amp, analog-to-digital converters, signal processors, digital-to-analog converters (DACs), power supplies, and audio amplifiers—and of course, the speakers themselves.
However, it takes a lot more to achieve an audiophile-level sound-reproduction experience in a vehicle than it does in a controlled environment at home. The in-cabin acoustics vary with occupancy, while road noise and other ambient noise sources overwhelm classic audiophile techniques. These factors put the spotlight on signal-processing techniques to get as close as possible to achieving high-end audio at reasonable cost.
The techniques used to accomplish this have already been widely adopted for home audio equipment. To get good-enough audio, speakers and equipment costing thousands of dollars used to be placed strategically about a specially prepared, acoustically neutral environment. Now, an amplifier’s built-in microphones and signal-processing algorithms can analyze room acoustics and compensate for a speaker system’s acoustic discoloration. This has made for exhilarating audio experiences in the home at ever-decreasing price points, even with single-speaker audio streaming devices from the likes of Sonos and Bose (and now Google, Apple, and Amazon).
However, in-vehicle audio playback needs more attention and processing horsepower. For example, acoustic noise cancellation (ANC) techniques are critical if road and engine noise are not to interfere with the audio experience. A typical ANC system is theoretically simple: detect the offending audio wave and send back a wave that’s 180 deg. out of phase through the vehicle’s speakers. Then cancel it out using destructive interference. This requires input mics, signal conditioning, continuously upgradable signal-processing algorithms, and a solid hardware implementation that can adapt to needs over time. It also must readily integrate with the audio codec and speaker system (Fig. 1).
The implementation sounds straightforward in theory and can be accomplished using, for example, least-mean-square (LMS) filters on an ARM Cortex-A15 running at 600 MHz. Yet it’s actually a complex endeavor that requires constant updating and re-assessment of algorithms.
This re-assessment and algorithm adoption may at one time have proven difficult. However, the incorporation of communications technologies into automotive multimedia systems means that ongoing updates to ANC algorithms are feasible using over-the-air update techniques. This tunability also applies to acoustic echo cancellation and equalization techniques.
The adoption of ANC techniques addressed one of the most annoying and difficult-to-address sources of noise: low-frequency engine-cylinder noise, especially on four-cylinder engines. With the increasing adoption of electric (EVs) and hybrid-electric vehicles (HEVs), engine noise will become less of an issue. But a number of other trends will make signal processing even more critical.
For example, advanced communications now requires the processing and storage of single-bit-stream high-resolution audio (HRA) from the output of oversampling sigma-delta converters that are oversampling at 128 fs and 256 fs.
From a user interface perspective, the integration of speech detection and recognition for voice command-and-control is overlapping with the need for ANC in the drive to enhance the user experience. Interestingly, the default method for automatic speech recognition (ASR) for human-machine communication (HMC) has been to use microphones for audio pickup. But now, companies like VocalZoom have adapted laser interferometer technologies that use lasers and advanced signal processing to extract voice commands from minute vibrations in the facial structure (Fig. 2).
In this adaptation of the interferometry, the laser and sensor are placed close to the driver, either in the rear view mirror or dashboard. This approach to voice-command detection is distinguished by its specificity to the driver and its immunity to ambient sounds or noise. It’s applicable to all classic applications of voice-activated command and control, including entertainment and communication systems, navigation, and voice-controlled driver assistance.
VocalZoom’s technology does not overshadow the enormous advances of acoustic microelectromechanical systems (MEMS) from companies such as Knowles. But it does underscore the continuing enhancement of voice-based interaction algorithms and underlying platforms.
Other applications that are putting demands on signal-processing capabilities include voice-based song retrieval, based on the user humming or singing a tune; speech synthesis for automobile two-way interaction; and IP-based audio for in-car networks. All of these features come on top of classic enhancements, such as the number of audio channels and the quality of each.
Keeping a Lid on Costs
Even as audio and signal processing improves, it would seem that fundamental switching elements like Class D amplifiers are sitting still. But that’s not the case. In fact, along with being more efficient than their Class A, B, or AB counterparts, Class D amplifiers can be used to enable speaker load diagnostics (Fig. 3).
In this context, speaker load diagnostics mean the detection of whether or not a speaker has been connected properly, and if there are any shorts to the vehicle chassis or battery lines. This ability to run speaker load diagnostics is becoming increasingly important as the number of speakers increases along with the overall wiring density in a typical vehicle. Ensuring that speakers are connected correctly before leaving a factory and being able to run diagnostics if a fault occurs in the field save time and cost with respect to both vehicle manufacturing and maintenance.
The typical scenario can implement both DC (woofer) and AC (tweeter) speaker diagnostics. AC is more complex, as it measures both the load impedance and phase of the test-signal frequency. The test setup shown in Fig. 3 is for detecting short-to-power and short-to-ground conditions. To perform the test, close switch 1 and 2 and open switch 3. This forces the OUTPUT N and OUTPUT P to the one half of the VDD supply (or 1.65 V in this case).
This information is sent to the ADC and the difference between the OUTPUT N and the 1.65 V is calculated, along with the difference between OUTPUT P and the 1.65 V. A report is then sent if a short is detected.
In summary, readily available signal-processing hardware and software techniques allow the integration of high-end audio playback. These approaches are now being combined with advanced in-vehicle man-machine interfaces and rapid diagnostics.