SARA-G3 and SARA-U2 series - System Integration Manual
UBX-13000995 - R26
System description
Page 72 of 217
1.10.3
Voice-band processing system
1.10.3.1
SARA-G340 / SARA-G350 modules audio processing
The voice-band processing on the SARA-G340 / SARA-G350 modules is implemented in the DSP core inside the
baseband chipset. The analog audio front-end of the chipset is connected to the digital system through 16 bit
ADC converters in the uplink path, and through 16 bit DAC converters in the downlink path. External digital
audio devices can directly be interfaced to the DSP digital processing part via the I
2
S digital interface. The analog
amplifiers are skipped in this case.
The voice-band processing system can be split up into three different blocks:
Sample-based Voice-band Processing (single sample processed at 8 kHz, every 125 µs)
Frame-based Voice-band Processing (frames of 160 samples are processed every 20 ms)
MIDI synthesizer running at 47.6 kHz
These three blocks are connected by buffers and sample rate converters (for 8 to 47.6 kHz conversion)
I2S_RXD
Switch
MIC
Uplink
Analog Gain
Uplink
Filter 2
Uplink
Filter 1
To
Radio TX
Uplink
Digital Gain
Downlink
Filter 1
Downlink
Filter 2
MIDI
Player
SPK
Switch
I2Sx TX
I2S_TXD
Scal_Rec
Digital Gain
SPK
Analog Gain
Gain_Out
Digital Gain
From
Radio RX
Speech
Level
I2Sx RX
Sample Based Processing
Frame Based Processing
C
ir
cu
la
r
B
u
ff
er
Sidetone
Digital Gain
DAC
ADC
Tone
Generator
AMR
Player
Hands-Free
V
o
ic
eb
an
d
S
am
p
le
B
u
ff
er
Figure 32: SARA-G340 / SARA-G350 modules audio processing system block diagram
The sample-based voice-band processing main task is to transfer the voice-band samples from either analog
audio front-end uplink path or I2Sx RX path to the Voice-band Sample Buffer and from the Voice-band Sample
Buffer to the analog audio front-end downlink path and/or I2Sx TX path. While doing this the samples are scaled
by digital gains and processed by digital filters both in the uplink and downlink direction and the sidetone is
generated mixing scaled uplink samples to the downlink samples (see the
u-blox AT Commands Manual
+UUBF, +UDBF, +UMGC, +USGC, +USTN commands).
The frame-based voice-band processing implements the Hands-Free algorithm. This consists of the Echo
Canceller, the Automatic Gain Control and the Noise Suppressor. Hands-Free algorithm acts on the uplink signal
only. Algorithms are configurable with AT commands (see the
u-blox AT Commands Manual
[3], +UHFP
command). The frame-based voice-band processing also implements an AMR player. The speech uplink path
final block before radio transmission is the speech encoder. Symmetrically, on downlink path, the starting block
is the speech decoder which extracts speech signal from the radio receiver.
The circular buffer is a 3000 word buffer to store and mix the voice-band samples from a Midi synthesizer. The
buffer has a circular structure, so that when the write pointer reaches the end of the buffer, it is wrapped to the
begin address of the buffer.
