GA3284 Datasheet, PDF (8/15 Pages) ON Semiconductor – Pre-configured DSP System

INSPIRIA GA3284
the greater the amount of attenuation. Simultaneously, in
each band, the masking threshold variations resulting from
the energy in other adjacent bands is taken into account.
Finally, the noise reduction gain is also adjusted to take
advantage of the natural masking of ‘noisy’ bands by speech
bands over time.
Based on this approach, only enough attenuation is
applied to bring the energy in each ‘noisy’ band to just below
the masking threshold. This prevents excessive amounts of
attenuation from being applied and thereby reduces
unwanted artifacts and audio distortion. The Noise
Reduction algorithm efficiently removes a wide variety of
types of noise, while retaining natural speech quality and
level.
Adaptive Directional Microphone
ON Semiconductor’s Adaptive Directional Microphone
(ADM) algorithm is a two−microphone processing scheme
for hearing aids. It is designed to automatically reduce the
level of sound sources that originate from behind or the side
of the hearing−aid wearer without affecting sounds from the
front. The algorithm accomplishes this by adjusting the null
in the microphone polar pattern to minimize the noise level
at the output of the ADM. The discrimination between
desired signal and noise is based entirely on the direction of
arrival with respect to the hearing aid: sounds from the front
hemisphere are passed unattenuated whereas sounds
arriving from the rear hemisphere are reduced.
The angular location of the null in the microphone polar
pattern is continuously variable over a range of 90 to 180
degrees where 0 degrees represents the front.
The location of the null in the microphone pattern is
influenced by the nature of the acoustic signals (spectral
content, direction of arrival) as well as the acoustical
characteristics of the room. The ADM algorithm steers a
single, broadband null to a location that minimizes the
output noise power. If a specific noise signal has frequency
components that are dominant, then these will have a larger
influence on the null location than a weaker signal at a
different location. In addition, the position of the null is
affected by acoustic reflections. The presence of an acoustic
reflection may cause a noise source to appear as if it
originates at a location other than the true location. In this
case, the ADM algorithm chooses a compromise null
location that minimizes the level of noise at the ADM
output.
FrontWave Directionality
The FrontWave block provides the resources necessary to
implement directional microphone processing. The block
accepts inputs from both a front and rear microphone and
provides a synthesized directional microphone signal as its
output. The directional microphone output is obtained by
delaying the rear microphone signal and subtracting it from
the front microphone signal. Various microphone response
patterns can be obtained by adjusting the time delay.
The FrontWave circuit also provides a fixed filter for
compensating the sensitivity and frequency response
differences between microphones. The filter parameters are
adjusted during product calibration.
A dedicated biquad filter following the FrontWave block
has been allocated for low frequency equalization to
compensate for the 6 dB/octave roll−off in frequency
response that occurs in directional mode. The amount of low
frequency equalization that is applied can be determined
during product calibration.
ON Semiconductor recommends using matched
microphones with FrontWave, although calibration is fully
possible using unmatched microphones.
A/D and D/A Converters
The system’s two A/D converters are second order
sigma−delta modulators operating at a 2.048 MHz sample
rate. The system’s two audio inputs are pre−conditioned
with antialias filtering and programmable gain
pre−amplifiers. These analog outputs are over−sampled and
modulated to produce two, 1−bit Pulse Density Modulated
(PDM) data streams. The digital PDM data is then
decimated down to Pulse−Code Modulated (PCM) digital
words at the system sampling rate of 32 kHz.
The D/A is comprised of a digital, third order sigma−delta
modulator and an H−bridge. The modulator accepts PCM
audio data from the DSP path and converts it into a 64−times
or 128−times over−sampled, 1−bit PDM data stream, which
is then supplied to the H−bridge. The H−bridge is a
specialized CMOS output driver used to convert the 1−bit
data stream into a low−impedance, differential output
voltage waveform suitable for driving zero−biased hearing
aid receivers.
HRX Head Room Expander
The Inspiria GA3284 has an enhanced Head Room
Expander (HRX) circuit that increases the input dynamic
range of the Inspiria GA3284 without any audible artifacts.
This is accomplished by dynamically adjusting the
pre−amplifier’s gain and the post−A/D attenuation
depending on the input level.
Channel Processing
Figure 6 represents the I/O characteristic of independent
AGC channel processing. The I/O curve can be divided into
the following main regions:
• Low input level expansion (squelch) region
• Low input level linear region
• Compression region
• High input level linear region (return to linear)
www.onsemi.com
8