RHYTHM R3910
www.onsemi.com
9
RHYTHM R3910 OVERVIEW
R3910 is a programmable multi−processor DSP platform
implemented on a thin−stacked package. This DSP platform
is the hearing industry’s first 90 nm Silicon−on−Chip
platform enabling design of highly−efficient and flexible
hearing aid solutions. The multi−processor DSP system
maximizing MIPS/mW with a unique reconfigurable
architecture, integrated high−resolution dual ADC and a
single DAC available in miniaturized package sizes,
offering unmatched DSP processing capability and
flexibility in an ultra small footprint with best in the industry
power consumption. R3910 incorporates industry leading
hearing algorithms allowing for easy integration into a wide
range of hearing products.
The DSP core implements FrontWave directional
processing, programmable filters, adaptive algorithms,
compression, wideband gain, and volume control. The
adaptive algorithms include Adaptive Noise Reduction,
Adaptive Feedback Cancellation and Automatic Adaptive
Directional Microphones.
Adaptive Noise Reduction reduces audible noise in a low
distortion manner while preserving perceived speech levels.
The Adaptive Feedback Canceller reduces acoustic
feedback while offering robust performance against pure
tones. The Adaptive Directional Microphone algorithm
automatically reduces the level of sound sources that
originate from behind or from the side of the hearing−aid
wearer without affecting sounds from the front.
Additionally, the Automatic Adaptive Directional
Microphones algorithm automatically reduces current by
turning off the second input channel if it is not needed.
The iLog 4.0 Datalogging feature records various
parameters every 4 seconds to 60 minutes (programmable)
during use of the device. Once these parameter values are
read from the device, they can be used to counsel the user and
fine tune the fitting.
iSceneDetect 1.0 is an classification algorithm that senses
the users environment and automatically optimizes the
hearing aid to maximize user comfort and audibility in that
environment without any user interaction. R3910 supports
iSceneDetect in 1 mic omni, static directional or adaptive
directional modes.
R3910 comes with Evoke advanced acoustic indicators.
Evoke allows manufacturers to provide more complex,
multi−frequency tones, in addition to traditional
programmable tones for memory changes and low battery
indication, which can simulate musical notes or chords.
R3910 is equipped with a noise source that can be used in
treating tinnitus. The Tinnitus Treatment noise can be
shaped and attenuated and then summed into the audio path
either before or after the volume control.
The Narrow−band Noise Stimulus feature allows the user
to generate stimuli from the device that can be used for in situ
audiometry. R3910 delivers advanced features and
enhanced performance previously unavailable to a product
in its class. As well, R3910 contains security features to
protect clients’ intellectual property against device cloning
and software piracy.
SIGNAL PATH
There are two main audio input signal paths. The first path
contains the front microphone and the second path contains
the rear microphone, telecoil or direct audio input as selected
by a programmable MUX. The front microphone input is
intended as the main microphone audio input for single
microphone applications.
Analog input signals should be ground referenced to
MGND (microphones, telecoils, DAI). MGND is internally
connected to GND to minimize noise, and should not be
connected to any external ground point.
In iSceneDetect, FrontWave, ADM or Automatic ADM
operation, a multi−microphone signal is used to produce a
directional hearing aid response. The two audio inputs are
buffered, sampled and converted into digital form using dual
A/D converters. The digital outputs are converted into a 32
kHz or 16 kHz, 20−bit digital audio signal. Further IIR filter
blocks process the front microphone and rear microphone
signals. One biquad filter is used to match the rear
microphone’s gain to that of the front microphone. After
that, other filtering is used to provide an adjustable group
delay to create the desired polar response pattern during the
calibration process. In iSceneDetect, ADM and Automatic
ADM, the two microphone inputs are combined in an
adaptive way while in FrontWave operation the combination
is static.
In the telecoil mode gains are trimmed during Cal/Config
process to compensate for microphone/telecoil mismatches.
The FrontWave block is followed by four cascaded biquad
filters: pre1, pre2, pre3 and pre4. These filters can be used
for frequency response shaping before the signal goes
through channel and adaptive processing.
The channel and adaptive processing consists of the
following:
• Frequency band analysis
• 1, 2, 4, 6 or 8 channel WDRC
• 16 frequency shaping bands (spaced linearly at 500 Hz
intervals, except for first and last bands)
• 128 frequency band adaptive noise reduction
• Frequency band synthesis
After the processing the signal goes through two more
biquad filters, post1 and post2, which are followed by the
AGC−O block. The AGC−O block incorporates the
wideband gain and the volume control. There are also two
