Audio Processing Solutions

Asia Analog Tech Day August 2009
Audio Processing Solutions
By Dominik Hartl and Yang Boon Quek 郭嚴文
d-hartl@ti.com quekyb@ti.com
Audio Applications Engineering
Audio and Imaging Products

Topics
• DRC / AGC
• 3-D
• bass boost
• echo cancellation
• noise cancellation
Relevant Products: TAS57xx, TPA2016/7, TAS550x,
TAS33xx, TSC2117, AIC3xxx

DRC and AGC

Overview
• DRC = Dynamic Range Compression
• AGC = Automatic Gain Control
• Detects the input signal envelope
• Varies the gain automatically
– Boosts up “softer” sound
– Reduces very loud signals to prevent clipping

TAS5706 DRC

TAS5706 DRC Fixed Gain Region
-75dBV
• Gain is constant in
this region
• Usually used to
boost low signals
-70dBV

TAS5706 DRC Compression Region
-12dBV
• Gain is decreasing
or increasing in this
region
-40dBV
• Usually used to
compress large
signals

TPA2016D2 AGC Output Waveform
• Optimized for enhancing the perceived audio
loudness (eg. Movies in noisy enviroments)
Without TPA2016D2:
With TPA2016D2:

TPA2016 Automatic Gain Control
Note that AGC do not
work below the 1:1 line.
i.e. min gain is 0dB
• Looks different
but is actually
almost the
same as DRC
• AGC has added
features
- Limiter
- Noise Gate
- Rotation point
fixed at 10dBV
- more flexibility

Limiter Level
• Sets max output voltage level
• Below max speaker power rating
• Below min supply voltage
VOUT
(dBV)
Limiter Level = 630mW
Limiter Level = 500mW
Limiter Level = 400mW
1:1
Increasing
Limiter
Level
Decreasing
Limiter
Level
VIN (dBV)
Fig. Varying Limiter Levels
VOUT
(dBV)
Large Fixed Gain
Small
Fixed
Gain
1:1
VIN (dBV)
Fig. Variation with Fixed Gain

Attack, Release and Hold Time
• Attack time: Min time between gain decreases
• Release time: Min time between gain increases
• Hold time: Min time between attack and release
• It is recommended to use default times
Input Signal
Amplitue (Vrms)
Gain (dB)
Attack time
Release time
Hold time
Time end
Time reset
Hold timer not used after
first gain increase
time

Noise Gate
• Prevents AGC from changing the gain when
there is no audio at the input of the amplifier
• Avoid amplifying noise floor

TPA2016 Automatic Gain Control
• Visualize your input vs output relationship in
your design

TAS5709/10 - 2 Band DRC
• 2 DRC controls.
• DRC1 controls the high freq audio.
• DRC2 controls the low freq audio.
• Note that you can change the audio cutoff freq
that DRC1 and DRC2 controls.
• Compress bass frequencies more aggressively
than mids/highs to prevent “cabinet” rattle.

L_IN
TAS5709/10 - 2 Band DRC
7 EQ BQs
Split the bass out
from the mid/highs
R_IN
7 EQ BQs
BQ1
High Pass
Filter
BQ1
Low Pass
Filter
BQ1
High Pass
Filter
BQ1
Low Pass
Filter
• Independent thresholds for Sub & Mid/High
frequencies
• Compress bass frequencies more aggressively
than mids/highs to prevent “cabinet” rattle.
• Prevent speaker damage
DRC-1
DRC-2
DRC-1
DRC-2
L_OUT
R_OUT
On-Chip DAP
Dampen the bass more
aggressively to prevent
cabinet rattle
• Audio Processing Supported
• 2-Band DRC + Speaker EQ (7 BQ per L/R)

d
B
V
+15
+12.5
+10
+7.5
+5
+2.5
+0
-2.5
-5
-7.5
-10
-12.5
-15
-17.5
-20
-22.5
-25
-27.5
Lab Results
TAS5709 - 2 band DRC
-30
-40 -37.5 -35 -32.5 -30 -27.5 -25 -22.5 -20 -17.5 -15 -12.5 -10 -7.5 -5 -2.5 +0
Sweep Trace Color Line Style Thick Data Axis Comment
1 1 Blue Solid 1 Anlr.Level A Left DRC1=No DRC. DRC2= No DRC
2 1 Magenta Solid 1 Anlr.Level A Left DRC1=Ratio 4, Threshold -25dB, offset=0. DRC2 = No DRC
3 1 Yellow Dash Dot 1 Anlr.Level A Left DRC1= No DRC; DRC2=Ratio 4, Threshold -25dB, offset=0.
Input = 2kHz. No effect if only DRC2 is enabled, since input is High Freq.
dBFS

d
B
V
+15
+12.5
+10
+7.5
+5
+2.5
+0
-2.5
-5
-7.5
-10
-12.5
-15
-17.5
-20
-22.5
-25
-27.5
Lab Results
TAS5709 - 2 band DRC
-30
-40 -37.5 -35 -32.5 -30 -27.5 -25 -22.5 -20 -17.5 -15 -12.5 -10 -7.5 -5 -2.5 +0
Sweep Trace Color Line Style Thick Data Axis Comment
1 1 Blue Solid 1 Anlr.Level A Left DRC1=none, DRC2=none
2 1 Magenta Solid 1 Anlr.Level A Left DRC1=none, DRC2=ratio4, threshold = -25dB, offset=0dB
3 1 Yellow Dash Dot 1 Anlr.Level A Left DRC1=ratio4, threshold = -25dB, offset=0dB; DRC2=none
Input = 200Hz. DRC1 has no effect, since input is Low Freq.
dBFS

3D Surround

Overview
• Supported Products : TAS5716, TAS5709/10, AIC3xxx
• Basic formula for implementing 3D Surround
L3D R3D L hpf
R hpf
1 { 2
1 { 2
( R L)}
( L R)}
• Low-frequency energy in the music of most stereo audio
tracks is monaural (i.e., left and right channels are the
same) and non-directional.
• We can easily vary the amount of 3D effects
a L b
hpf ( R L)}
L3D R3D
c R d hpf
1 { 2
1 { 2
( L R)}

L
R
TAS5709 3D
On-Chip DAP
DELAY
BQ
DELAY
BQ
x (+0.5)
x (-0.5)
1 st HPF
Fc:180Hz
EQ:
Fc:1kHz, G:3dB
BW: 2kHz
BQ BQ
Side Signal: S = (L-R)/2 3D Filter: Filter the Bass
out and boost the mids
• Enhanced Audio Experience with soundstage widening
• 3D algorithm can be fine tuned. The mixers, Biquad
filters, etc. are programmable.
x 1
x 1
DELAY
BQ
DELAY
BQ
x (+0.75)
x (+1.25)
x (-1.25)
x (+0.75)
L‟
R‟
7 EQ BQs
Add the Side signal
to the Left / Subtract
It from the Right
7 EQ BQs
• Audio Processing Supported
• 3D Effects + Speaker EQ (7 BQ per L/R) + Bass
Boost

TAS5709 3D
• The mixers, Biquad filters registers, etc. are
programmable through I2C.

Example: AIC3111 3D
Check out the demo after the presentation!

Bass Boost

Overview
• Provide improved bass performance without the use of
subwoofer speakers
• Basic formula for implementing Bass Boost
Lout
L
a L b
DRC{
BBF(
2
L
Rout
a R b
DRC{
BBF(
2
where BBF is LPF + EQ
• DRC is needed for loudness control to prevent excessive
clipping in low frequency contents
• We can easily vary the amount of Bass Boost effects
R
R
)}
)}

L
R
TAS57xx Architecture
(L+R)/2
BQ
2 nd HPF
Fc:200Hz
x (+0.5)
x (+0.5)
2 nd HPF
Fc:200Hz
BQ
2 nd LPF
Fc:200Hz
• Improved bass performance without subwoofer speakers
• DRC protects speaker from damage/clipping
BQ BQ
• Bass Boost algorithm can be fine tuned / adjusted. The
mixers, Biquad filters, etc. are programmable.
.
On-Chip DAP
EQ:
Fc:180Hz, G:12dB
BW: 300Hz
Bass Boost Filter
DRC
K: 1:30
O:0dB
T:-16dB
Energy: 1ms
Attach: 100ms
Decay: 10000ms
x (+1)
x (+2.4)
x (+2.4)
x (+1)
• Audio Processing Supported
• Bass Boost+3D Effects+Speaker EQ (7 BQ per L/R)
L’
R’
7 EQ BQs
7 EQ BQs

Noise Reduction

Overview
1. The input audio signal is split into multiple frequency
bands according to the user input.
The center frequency for each of the high frequency and low
frequency bandpass filters are variables controlled by the user.
2. Each frequency band is „gated‟ so that only signal that is
a certain threshold above the estimated noise floor is
passed.
Both the threshold and amount of attenuation level is user
programmable.
3. The output from each of the frequency bands is summed
to form the noise reduction output.

Noise Reduction Algorithm

Noise Reduction GUI
• Threshold: Defines the level above the estimated noise floor the signal must be (within it‟s band) to be
considered „signal‟ as opposed to „noise‟. It spans from 0 dB to 20 dB and is used along with the Attenuation property
to differentiate between „signal‟ and „noise‟.
• Attenuation: Defines how much noise attenuation will be applied. It spans from 0 dB to 20 dB and is used
along with the Threshold property todifferentiate between „signal‟ and „noise‟.
• Low Freq & High Freq: Define the bandwidth over which the noise reduction filter bank is to be applied.
The “Low Freq” control spans from 100 Hz to 1000 Hz. The “High Freq” control spans from 1000 Hz to 18,472.5 Hz.
Available for:
AIC3254
AIC3111
AIC36
TSC2117

Echo Cancellation

Overview

Components
• NLMS Adaptive Filter:
This filter is based on the Normalized Least Mean Square (NLMS) algorithm. The NLMS
algorithm incorporates an iterative procedure designed to make successive corrections
to its weighted filter coefficients based on the measured error. This procedure eventually
leads to the minimization of the mean squared error. The error in this application is the
difference in the echo at the uplink (Tx) input and the estimated echo calculated by the
NLMS algorithm.
• NLP Controller:
When the NLMS algorithm cannot completely remove echo, the NLP controller will
suppress any remaining residual echo. The NLP controller initially classifies the
communication scenario and then applies an appropriate amount of attenuation to the
Uplink (Tx) and Downlink (Rx) signals to make the residual echo less audible.

Components
• Center Clipper (CC):
A Center Clipper (CC) module is applied to replace small-amplitude audio samples with
zeros to make low-level residual echo completely inaudible.
• Comfort Noise Generator (CNG)
The CNG generates pink noise to make background noise sound smoother in the uplink
(Tx). The CNG automatically estimates the near-end background noise level and
generates pink noise that matches the same level as the estimated background
noise.The comfort noise generator may mask low level transients generated by the NLP
module when changing modes as well as avoid the effects of “dead air” caused by the
elimination of all residual noise on the line.

AIC3254 implementation

Thank You!
Questions?