Self-Calibrating Audio Signal Equalization

1
Self-Calibrating Audio Signal Equalization

• Greg Burns
• Wade Lindsey
• Kevin McLanahan
• Jack Samet

2
Project Scope

• In any closed room, standing waves exist that
  change the way audio signals arrive at the ear.
• Variations in amplifier design, speaker
  efficiency, and room geometry affect the
  frequency response, degrading it from flat-band
  operation.
• The goal of this project is to automatically
  calibrate an audio signal to compensate for these
  effects.

3
Audio Fundamentals

• Pink Noise is a randomly generated signal that
  exhibits a constant voltage per octave.
• A spectrum analyzer can be used to obtain the
  actual frequency response of an audio signal when
  placed in a test position in a room.
• A graphic equalizer can then be used to adjust
  the amplifier input to compensate for any
  deviations off flat-band response

4
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
MUX
Spectrum Analyzer
Equalizer
Audio Preamp
5
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
Equalizer
MUX
Spectrum Analyzer
Audio Preamp
6
Equalizer Specifications

• 10 Bands (32, 64, 128, 256, 512, 1024, 2048,
  4096, 8192, 16384 Hz)
• Filters, input, and output constructed using
  LM351 op-amps
• Discrete components and 10kO DS1803 digital
  potentiometers

7
10-Band Equalizer Circuit
8
Built Equalizer
9
Equalizer Response

• Frequency response of equalizer with varying
  resistances tested with HP VEE.

10
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
MUX
Spectrum Analyzer
Equalizer
Audio Preamp
11
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
MUX
Spectrum Analyzer
Equalizer
Audio Preamp
12
Pink Noise Specifications

• Creates pseudorandom digital noise for white
  noise in first stage at 3dB/dec
• Second stage pink noise filter at -3dB/dec
• Frequency response 20 Hz 20 kHz
• 33-bit resolution in shift register for
  pseudorandom number generation
• Line level output at 150mV rms

13
Pink Noise Generation

• Equal voltage per octave across audio band.

• FFT of Pink Noise viewed on oscilloscope.

14
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
MUX
Spectrum Analyzer
Equalizer
Audio Preamp
15
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
MUX
Spectrum Analyzer
Equalizer
Audio Preamp
16
Spectrum Analyzer

• Samples input signal from microphone
• Performs an FFT (Fast Fourier Transform)
  algorithm to extract frequency components
• Compares relative frequency levels to optimal
  flat-band response
• Samples microphone input at 19.2 µs
• Sample length of 256 data points at 8-bit
  resolution

17
FFT Explained

• Implementing Cooley/Tukey FFT algorithm.
• Has Big O of N log N
• Takes Fourier matrix of power 2 (28 in our case)
• Breaks into 2 log N matrices and performs
  multiplications on roots of unity (e2pihk/N)
• Ultimate result returns a vector with frequency,
  phase, and magnitude information.

18
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
MUX
Spectrum Analyzer
Equalizer
Audio Preamp
19
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
MUX
Spectrum Analyzer
Equalizer
Audio Preamp
20
Microcontroller Software
Initialization Mode
Sets all initial variables and default settings.
Outputs pink noise signal through speakers,
receives spectral data from analyzer, and adjusts
equalizer to compensate.
Calibration Mode
Operation Mode
Resets MUX to audio source and selects current
room location to compensate.
21
Initialization Mode

• Sets system timers and interrupts
• Configures input and output pins
• Defaults variables to initial conditions
• Initializes I2C transfers
• Sets digital pots to a predetermined ideal
  flat-band response

22
Calibration Mode

• Gathers data from spectrum analyzer output
• Compares current frequency response peaks to
  ideal response
• Adjusts digital potentiometers based upon
  previous comparison
• Repeats until current frequency response and
  ideal response fall within 5 tolerance
• Calibrates for every room position

23
Operation Mode

• Switches audio source from pink noise to preamp
• Sets digital pots to specific values
  corresponding to room location
• Monitors Sensor Arrays for room location
  variations

24
Block Diagram
Sensor Array
PIC Microcontroller
Pink Noise Generator
MUX
Spectrum Analyzer
Equalizer
Audio Preamp
25
Future Improvements

• Increased number of frequency bands on EQ
• Use of DSP processor for improved FFT performance
• Use of audio-grade tolerance components
• Allow for wide variety of sensor array
  configurations

26
Conclusions

• Economically feasible and marketable
• Modular design allows for easy implementation,
  innovation, and reproduction
• Compatible with most modern stereo systems
• Overall a universally usable product from
  personal to commercial applications

27
Questions?
28
Fin