The tension between cost and quality is one of the fundamental tradeoffs in the design of consumer electronics devices—and many other systems. Customers predominantly select among competing products based on price, especially in these challenging economic times, but consumers are also unwilling to short-change perceived quality. For example, to minimize bill-of-materials costs, engineers prefer to incorporate low-cost speakers in their designs. These entry-level transducers typically exhibit notable shortcomings, however: frequency range deficiencies (especially at the low end of the audible spectrum), along with high distortion and low output volume, both resulting from poor power-handling capabilities.
Fortunately, digital signal pre-processing can significantly compensate for speakers' analog shortcomings. This was the task that BDTI's engineers undertook in partnership with a leading chip supplier, whose 32-bit microcontrollers integrate numerous digital signal processing features, including a floating-point unit, a multiply-accumulate unit, and barrel shifter. The supplier’s MCU evaluation board included an ultra-low-cost speaker, and the supplier wondered if the DSP features of the MCU could be used to improve sound quality.
BDTI's engineers analyzed the speaker's capabilities and concluded that a set of relatively simple, well-known audio pre-processing algorithms would enable significant improvement in audio quality. BDTI applied different techniques to different portions of the audio spectrum to clean up the output of the speaker without expending a large amount of processor resources.
Since BDTI’s engineers found that the speaker’s high-frequency performance was adequate, the BDTI algorithms passed high and upper-midrange audio data (down to around 800 Hz) through unchanged. Conversely, the speaker was judged incapable of adequately reproducing bass frequencies (ranging up to about 400 Hz), even via so-called "virtual bass" (a psychoacoustic effect also commonly known as the "missing fundamental" approach), so audio data in this range was discarded.
The lower-midrange portion of the spectrum received the bulk of audio processing attention. Dynamic range compression increased the volume during quiet passages and decreased it during loud segments. Soft clipping, applied with great care to avoid the audible effects of the harmonics and inter-modulation distortion it can potentially create, additionally amplified quiet signals. And a band-pass filter further boosted the lower-midrange signal, while also shaping harmonics generated by the soft clipping function to ensure a pleasing effect. Finally, the algorithm's limiter stage acted as a volume control, applying gain to the input signal to boost quiet audio content and automatically reduce the gain as necessary to prevent audible distortion.
The resultant sound quality was audibly superior to the unprocessed original, akin to that delivered by a more expensive speaker. The evaluation board now provides an effective demonstration of the MCU’s low-cost digital signal processing capabilities. And BDTI provided both source and executable software to the MCU supplier for distribution to its customers, along with detailed documentation, giving the supplier’s customers an easy starting point for building their own low-cost, real-time audio processing functions.
If you'd like to harness BDTI's extensive audio expertise in your next project, contact Jeremy Giddings at +1 925-954-1411 or giddings@BDTI.com.