CS48LV12/13 Ver la hoja de datos (PDF) - Cirrus Logic
Número de pieza
componentes Descripción
Lista de partido
CS48LV12/13 Datasheet PDF : 26 Pages
| |||
3.6 Acoustic Echo Cancellation (AEC)
adjustments to compensate for changes in device orientation. This results in better off-axis performance than do traditional
techniques with fixed acceptance angles. The noise-reduction block uses highly adaptive spatial analysis, spectral
analysis, and audio classification to quickly adjust for changes in audio environment and talker position.
Simple adjustments made through the RAPID2 tool allow the noise reduction block to be optimized for a variety of
microphone spacing and placements. Tightening standards on signal-to-noise-level improvement (SNRI) and voice quality
require careful balance to assure compliance and quality user experience. The RAPID2 tool provides the granularity of
control required to optimize user experience and compliance.
3.6 Acoustic Echo Cancellation (AEC)
In principle, the AEC works by recognizing the Rx signal that reappears, with some delay, in the Tx signal being processed
for transmission. Before transmission, the echo is removed by subtracting or canceling it out. In reality, canceling all
perceptible echo while maintaining full-duplex communication requires a complex set of monitoring control and digital DSP
algorithms. The CS48LV12/13’s AEC in combination with the Voice DSP Control and Detection group can provide
conference room–grade speakerphone performance within the performance constraints of the transducers used.
However, achieving high quality, full-duplex, echo-free performance in both handset and handsfree/speakerphone modes
requires conscientious system-level attention before completing component selection and design. Contact your Cirrus
Logic sales representative regarding available technical support early in the product-design process.
3.6.1 Residual Echo Suppressor/Nonlinear Processor
These blocks identify and suppress any nonlinear echo components not canceled by the AEC. The residual components
include distortion commonly introduced by small transducers, such as those used in mobile devices and is critical to
achieving a clean-sounding echo-free Tx voice stream.
3.7 Spectrally Matched Comfort Noise Generator (CNG)
Full-duplex communication can be classified into four possible states:
• Silence (neither the near- nor the far-end talker is speaking)
• Single-Talk Rx (only the far-end talker is speaking)
• Single-Talk Tx (only the near-end talker is speaking)
• Double-Talk (both talkers are speaking simultaneously)
During the Single-Talk Rx state, the AEC suppresses the Rx signal to eliminate all echo when no near-end voice is
detected. This can cause the person on the far end to perceive modulation between hearing ambient far-end noise and
then silence. To avoid this, the CNG samples ambient noise, synthesizes it, and injects it into the Tx stream whenever
there is silence, providing the far-end listener with a more natural, constant audio experience.
3.8 Dynamic PEQ
Cellular providers and industry standards require that Tx frequency response conforms to a tight envelope. Furthermore,
compensating for transducers characteristics and achieving natural voice character may require significant parametric
equalization of the Tx voice stream. This block provides up to four concurrent filters, each of which can be defined as one
of eight different filter types that can be combined to achieve exact frequency response requirement. Each filter has
tunable frequency, gain, and Q or bandwidth. The following filter types are available: low pass, high pass, low shelf, high
shelf, band pass, peaking, notch, and all pass.
In addition to functioning as a traditional parametric equalizer, this block can be operated in Ambient Aware Mode, where
EQ settings are automatically adjusted in real time, based on either loudness or noise levels. This feature can be used to
improve intelligibility under noisy conditions or maintain natural sound over changing loudness levels.
The dynamic PEQ block is the same for both the Tx and Rx voice DSP chains.
9
DS1057F1
Share Link: 