Dolby AC-3 Audio Encoding

1
Dolby AC-3 Audio EncodingTHX
Wai Kam (Winnie) Henele Adams Peter Boettcher
2
Multichannel Audio

• Audio used in film and various entertainment
  ranges from mono to stereo to 5.1 system
• Wideband Audio 48 KHz 16 bits 6 Channels
  46.08 Mbits/sec

3
AC-3 Digital Audio Coding

• Flexible Multi-channel coding encode up to 6
  channels and downmix according to the listeners
  available resources
• Current Standard for HDTV audio compression also
  used in LD, DVD
• Sample rates 32 KHz, 44.1KHz, 48KHz
• Data rates 32 Kbps to 640 Kbps
• 1st AC-3 coded film Star Trek VI in Dec. 91

4
AC-3 Encoder

• Frequency Domain Transform
• 512 point MDCT with 50 overlap
• 2 x 256 point MDCT in transient signal
• Bit allocation in floating point representation
• Mantissa and Exponent are represented differently
  
• Bits of Mantissa are adaptively allocated

5
AC-3 Decoder
1st Step
Recover Mantissa according to Psychoacoustic
model, inverse transform and de-mux to different
output channels
Recover PCM values for each channel
2nd Step
6
Bit-stream Syntax
1 Frame represents 1536 PCM samples for all
channels
Frame Structure
SISync. Info BSIBitstream Info CRC for error
correction Aux Data for private control
1 Block represents 256 PCM for each channel
Block Structure
7
Encoding Philosophy

• Encoding Goals
• Minimize perceived sound quality degradation
• Keep the bit-rate within set value
• Bit allocation strategies
• forward-adaptive
• backward-adaptive
• parametric bit allocation

8
Forward Adaptive
9
Backward Adaptive
10
Parametric Bit Allocation

• Hybrid of Forward and Backward strategies

11
Parametric Bit Allocation

• Core is Backward Adaptive
• Primary Modifications
• Psychoacoustic model parameter adjustment
• Delta Bit Allocation

12
Filter Banks

• Filter Bank construction
• Overlapping blocks of 512 windowed samples
• Block 10.66 msec long (_at_48kHz)
• Adjustments for transients

13
Spectral Envelope

• Coefficients coded into an exponent and a
  mantissa
• Allows wide dynamic range
• Precision limitations in mantissa introduce noise
• Exponents characterize the spectrum
• Spectral envelope

14
Spectral Envelope

• Coding
• DC sent as absolute
• Exponents sent as differentials
• Groups of 3 encoded into 7 bit blocks
• Four exponent encoding strategies
• D15, D25, D45, and REUSE

15
Spectral Envelope
16
Bit Allocation

• Psychoacoustic Reductions
• Pool of bits for all channels
• Adjustments made to SNR
• Coping with Bit rate Conditions
• coupling

17
5.1 Channel Surround Sound

• 5 main channels - L, C, R, LS, RS
• 1 LFE (low-frequency effects) channel
• bandwidth is approximately 1 tenth, thus .1
• Unlike analog Dolby Surround, AC-3 5.1 format
  includes 6 discrete channels, but coded together
  for efficiency

18
Recommended Setup

• Left and Right stereo pair 45-60 degrees off
  center
• Surround speakers to side and slightly to rear
  (110 degrees)
• All speakers equidistant from listener, or use
  appropriate time delays in system.
• Subwoofer anywhere

19
Bass Management

• The subwoofer channel is NOT the same as the LFE
  channel
• LFE channel is created by the sound engineer at
  mixing time
• The subwoofer channel is produced by the decoder
  at playback time
• Dolby recommends caution in creating LFE channels
  at all

20
Use of the Channels

• Front channels
• Phantom center (like stereo), center channel
  alone, or both
• Surround channels
• Enhance depth and immersion
• Not well established
• LFE
• Only special bass effects

21
Decoder Downmixing
22
Decoder Downmixing

• Decoder adjusts playback for equipment
• 5.1 audio track can be mixed down to 4 channel
  surround or stereo
• Dynamic range reduction, LFE channel elimination

23
THX

• Two needs identified
• A need to correct the audible tonal and spatial
  errors caused by the playback of soundtracks
  designed in and for large theatres in the smaller
  environment of a home.
• A need to more accurately reproduce the complex
  and competing sound fields present in
  multi-channel sound playback.
• Plus, George Lucas wants to make money

24
THX Certification

• THX is like the Good Housekeeping Seal of
  Approval. It certifies that the product performs
  to a rigorously high standard that makes sense.
  THX provides the confidence and reassurance a
  consumer needs in a purchasing decision.
• -- Steve Baker, V.P. Sales Marketing, Denon
  Electronics

25
Questions

• What are the primary differences between MPEG
  audio and AC-3 coding?
• MPEG audio uses the forward adaptive bit
  allocation scheme whereas the AC-3 method uses
  the hybrid (parametric) scheme discussed earlier.
• The bitstream formats are different and allow
  different extensions and capabilities.