Audio Measurement

1
Audio Measurement

• Measurement, analysis and optimisation of large
  scale sound systems

2
Introduction Overview

• Professional audio systems intro
• Basic audio measurement theory
• System alignment key concepts
• Introduction to analysers
• Basic analyser measurements
• Example system analysis and optimisation

3
Professional Audio Systems
4
Mixing Consoles
5
Signal Processing
6
Amplifiers
7
Speaker Systems
8
Measuring Audio Systems

• From the perspective of audio system measurement,
  we focus on the objective goals
• The output of the mix console is the reference
  signal all the art happens before that
• And we measure the final signal in the acoustic
  domain, the result

9
Audio System Signal Flow
10
Audio System Signal Flow
11
Audio System Signal Flow
12
Audio System Signal Flow
13
Audio System Signal Flow
Ideally, almost every array element requires its
own level, delay and equalization. The
complexity of signal processing exponentially
increases as the size of array systems increase.
14
Audio System Signal Flow
There are subtle issues that must be addressed.
Have a look at how the elements in this array are
placed and curved in three dimensions. Its not
a trivial task to optimise this real world array.
15
System Optimisation

• This is our task, to manage this complex system
  of signal processing in order to produce the
  desired objective system performance at the
  listening position
• Whats in our tool kit to do the work?

16
Techniques of System Optimisation

• Verification of components
• Verification of installation
• Architectural modification
• Speaker positioning
• System delay settings
• Set system equalization
• Set system levels

17
Techniques of System Optimisation

• Verification of components
• Maximum output capability
• THD
• Noise
• Polarity
• Gain
• Frequency range
• Coverage pattern
• Proper operation

18
Techniques of System Optimisation

• Verification of installation
• Wiring polarity
• Wiring routing
• Wiring balanced/unbalanced

19
Techniques of System Optimisation

• Architectural Modification
• Locate major sources of reflection
• Examine options
• Fight with architects, scenic artists,
  preservationists and budget managers
• Absorb/diffract whatever you can with whatever
  you have

20
Techniques of System Optimisation

• Speaker Positioning
• On Axis analysis
• Vertical limits
• Horizontal limits
• Acoustic crossovers

21
Techniques of System Optimisation

• System delay setting
• Determine delay strategy
• Select position
• Find reference speaker arrival
• Set delay

22
Techniques of System Optimisation

• Set system equalization
• Identify single system response trends
• Room/EQ overlay techniques
• Verify result
• Combine subsystems
• Identify combined trends
• Update EQ

23
Techniques of System Optimisation

• Set system levels
• Calibrate measurement mics
• Set target dB SPL range
• Select representative on axis position
• Store master level reference trace
• Set level in subsystem on axis areas to match
  master level reference
• Optimise seam and overlap zones for maximum
  uniformity

24
Techniques of System Optimisation

• Verification of components
• Verification of installation
• Architectural modification
• Speaker positioning
• System delay settings
• Set system equalization
• Set system levels

25
Measurement Techniques
26
Measurement Techniques

• Currently there are two distinct methods of
  measurement commonly used in the design, analysis
  and optimisation of professional audio systems
• Logarithmic swept sine
• Source independent measurement

27
Logarithmic Swept Sine

• This relatively new measurement technique was
  introduced formally in Angelo Farinas AES paper
  in 2000 entitled,
• Simultaneous measurement of impulse response and
  distortion with a swept-sine technique

28
Logarithmic Swept Sine

• This new method supplants Maximum Length Sequence
  (MLS) measurement techniques
• Particularly in the area of large scale sound
  system design and optimisation

29
Logarithmic Swept Sine

• Log swept sine is favourable because
• Robustness against time variance
• You can use a large (high level) output signal
• Remote measurements can be performed without a
  loss in quality

30
Logarithmic Swept Sine

• Robustness against time variance
• Time variance happens in the real world due to
  air movement (wind), temperature and rigging (the
  array is never completely motionless)
• MLS techniques are delicate they fall over with
  any time variance

31
Logarithmic Swept Sine

• You can use a large output signal
• The Log Swept Sine technique isolates the linear
  response and non-linear response components in
  the measurement
• This means that the loudspeaker can be driven
  hard to overcome signal to noise ratio,
  environmental factors, etc.
• MLS techniques suffer when presented with even
  small non-linearities

32
Logarithmic Swept Sine

• Remote measurements can be performed without loss
  in quality
• Using two sources (e.g. a CD-player and a PC)
  that are not clocked together can be useful to
  avoid long cables from the PC to the loudspeaker
  and microphone.

33
Comparison of methods
34
Comparison of methods
35
Comparison of methods
36
Additional benefits
37
Logarithmic Swept Sine

• WinMLS is a commercially available product that
  performs Log Swept Sine analysis.

38
WinMLS example measurement
The Effects of Reflections
Time domain
Frequency domain
39
WinMLS example measurement
Subjective perceptions of reflections
40
WinMLS example measurement
Room impulse response
41
WinMLS example measurement
First 15m (44ms) of Impulse Response
Impulse from Monitor (direct sound)
Reflection off Mixing Desk
Time of Flight
Reflection off Ceiling
Reflection off Rear Wall
Reflection off Side Wall
42
WinMLS example measurement
Reflections in the frequency domain
Reflection off Ceiling
Reflection off Mixing Desk
Reflection off Side Wall
Monitor angled off axis to listening position
Reflection off Rear Wall
43
WinMLS example measurement
Active Monitor
Left
Sound Card
GENELECR
Line Output
Right
Optional loop-back
(for synchronisation)
Laptop Computer
AKG Phantom
Power Supply
Left
Microphone
Mic. Input
44
WinMLS

• For critical analysis and detailed design work,
  WinMLS provides an invaluable tool.

45
EASERA

• EASERA is also another measurement system that
  performs swept sine analysis

46
EASERA

• Multiple stimulus can be chosen for analysis

47
Measurement Techniques

• But what if there is an audience present? No one
  is willing to listen to logarithmically swept
  sine waves before or during a concert

48
WinMLS

• WinMLS is a Source-Dependent measurement
  technique.

49
Measurement Techniques

• Source Independent Measurement (SIM) techniques
  exist, and are very practical and applicable to
  large scale sound system analysis and optimisation

50
Source Independent Measurement

• Introduced formally in 1984 by John Meyer of
  Meyer Sound Laboratories, in an AES paper
  entitled,
• Equalization using voice and music as the source

51
Source Independent Measurement
1986 Based on the Hewlett-Packard 3582A
dual-channel FFT analyzer, the first generation
SIM.
52
Source Independent Measurement
1991 The SIM II FFT analyzer, released in 1991,
receives the prestigious RD 100 Award from RD
Magazine, for such innovations as its three
simultaneous transfer functions.
53
Source Independent Measurement
2003 SIM 3 launches, representing a major
advance, greatly increased power and added
features at a fraction of the original size and
cost.
54
Source Independent Measurement

• Measurement block diagram

55
Source Independent Measurement

• SIM is fundamentally based upon the Fast Fourier
  Transform
• Lets briefly review some FFT theory and
  fundamentals

56
Source Independent Measurement
57
Source Independent Measurement

• Why use FFTs?
• Can be used in dual-channel mode
• Contains complex response i.e. Magnitude and
  Phase
• Does not require a known source
• Best accuracy in a noisy environment via advanced
  averaging techniques

58
Source Independent Measurement

• Time Bandwidth Product
• FFT acquires a time record of data
• The lowest frequency measured has a period equal
  to the time record
• The lowest frequency measured is the bandwidth of
  the measurement
• The Time record (sec) x Bandwidth (Hz) 1. This
  is known as the Time/Bandwidth Product
• All frequency data points are multiples of the
  bandwidth

59
Source Independent Measurement
60
Source Independent Measurement
61
Source Independent Measurement

• Linear Frequency Axis
• FFT frequency computation is based on BANDWIDTH
  (Measured in Hz)
• Human hearing responds to PERCENTAGE BANDWIDTH
  (Measures in Octaves)
• FFTs compute low BW at Low frequencies and high
  BW at High frequencies
• Long time records are required to give good LF
  resolution - but give too much data at HF
• Solution Multiple Time Records of different
  lengths optimized for each Octave

62
Source Independent Measurement
63
Source Independent Measurement
64
Source Independent Measurement
Time records optimized for constant resolution
this is a piecewise approximation to a constant-Q
transform
65
Source Independent Measurement
66
Source Independent Measurement
67
Source Independent Measurement
68
Source Independent Measurement
69
Source Independent Measurement
70
Source Independent Measurement
71
Source Independent Measurement
72
Source Independent Measurement
73
Source Independent Measurement
74
Reading Phase Response
75
Reading Phase Response
76
Reading Phase Response
77
Phase Response Angle vs Frequency
78
Phase Response Angle vs Frequency
79
Low Pass Filters 1st to 4th order _at_ 1 kHz
80
Source Independent Measurement
81
SIM Averaging

• Why average?
• NOISE!
• Noise in the output creates errors in the
  transfer function data. Therefore any single
  sample with noise does not represent the best
  approximation of the device transfer function.
• Averaging improves the statistical validity of
  the data because random noise functions will
  divide out.

82
SIM FIFO Averaging
83
SIM FIFO Averaging
84
SIM FIFO Averaging
85
Coherence Noise in the measurement
86
Coherence Noise in the measurement
87
Factors affecting coherence
88
Coherence correlated noise
89
Coherence semi-correlated noise
90
Coherence uncorrelated noise
91
Coherence output is noise added to input
92
Coherence noise destabilizes the transfer
function
93
Coherence Thresholding
94
Advanced Optimization Techniques

• When designing an array, it is important to take
  significant contributing factors into account
• Low frequency coupling
• Environmental conditions
• These factors affect everything from how the
  system is wired up to how the system is equalized

95
Advanced Optimization Techniques
Wide Directivity _at_ Lower Frequencies
Narrow Directivity _at_ Higher Frequencies
96
Advanced Optimization Techniques
Wide Directivity _at_ Lower FrequenciesLow
Frequencies cover wide Area
Only Upper Speaker
Only Middle Speaker
Only Bottom Speaker
97
Advanced Optimization Techniques
Narrow Directivity _at_ Higher Frequencies High
Frequencies cover narrow Area
Only Upper Speaker
Only Middle Speaker
Only Bottom Speaker
98
Advanced Optimization Techniques
Frequency Response vs. Number of Speakers
Higher Frequencies didnt show increase in
relative Level
Lower Frequencies show increase in relative Level
Differences in Frequency Response using 1-2-4-8
Speakers
99
Advanced Optimization Techniques
Low Frequency Coupling must be corrected using
Equalization
The amount of Equalization is defined by-the
Amount of Speakers-the Length of the Array
The Length of the Array is defined by-the
Physical Size of the Speaker Cabinet
100
Advanced Optimization Techniques
Differences in Frequency Response using 1 vs 8
Milo before Corrective Equalization
LD-3 Controller parameters Model Milo Array
Size 8
Differences in Frequency Response using 1 vs 8
Milo after Corrective Equalization
101
Advanced Optimization Techniques
Environmental Conditions
8 Milo _at_ 16 kHz. Without Environmental Conditions
102
Advanced Optimization Techniques
Environmental Conditions
8 Milo _at_ 16 kHz. 20 Celsius, 50 Relative
Humidity
103
Advanced Optimization Techniques
8 Milo _at_ 4 kHz. Without Environmental Conditions
104
Advanced Optimization Techniques
8 Milo _at_ 4 kHz. 20 Celsius, 50 Relative Humidity
105
Advanced Optimization Techniques
8 Milo _at_ 1 kHz. Without Environmental Conditions
106
Advanced Optimization Techniques
8 Milo _at_ 1 kHz. 20 Celsius, 50 Relative Humidity
107
Advanced Optimization Techniques
Frequency Response at 16 m
Environmental Attenuation
Blue Color Without Environmental Conditions
Red Color _at_ 20Celsius, 50 Relative Humidity
108
Advanced Optimization Techniques
Frequency Response at 34 m
Environmental Attenuation
Blue Color Without Environmental Conditions
Red Color _at_ 20Celsius, 50 Relative Humidity
109
Advanced Optimization Techniques
Frequency Response at 61 m
Environmental Attenuation
Blue Color Without Environmental Conditions
Red Color _at_ 20Celsius, 50 Relative Humidity
110
Chart of Attenuation per meter _at_ 16 kHz
111
Attenuation Scenarios
Dry and Cold, produce Small Attenuation of Higher
Frecuencies Dry and Hot, produce High Attenuation
of Higher Frecuencies
Blue Color _at_ 10 Relative Humidity, 8Celsius
Red Color _at_ 10 Relative Humidity, 38Celsius
112
Attenuation Scenarios
Wet and Cold, produce High Attenuation of Higher
Frecuencies Wet and Hot, produce Small
Attenuation of Higher Frecuencies
Brown Color _at_ 80 Relative Humidity, 38Celsius
Green Color _at_ 80 Relative Humidity, 8Celsius
113
Attenuation Scenarios
Cold and Dry, produce Small Attenuation of Higher
Frecuencies Cold and Wet, produce High
Attenuation of Higher Frecuencies
Blue Color _at_ 8Celsius, 10 Relative Humidity
Green Color _at_ 8Celsius, 80 Relative Humidity
114
Attenuation Scenarios
Hot and Dry, produce High Attenuation of Higher
Frecuencies Hot and Wet, produce Small
Attenuation of Higher Frecuencies
Brown Color _at_ 38Celsius, 80 Relative Humidity
Red Color _at_ 38Celsius, 10 Relative Humidity
115
Advanced Optimization Techniques
Patching for High Frequency Correction
Throw per speaker 1 61m 2 55m 3 48m 4 42m 5
35m 6 28m 7 22m 8 16m
116
Advanced Optimization Techniques
Patching for High Frequency Correction
Short Throw 1 61m 2 55m 3 48m 4 42m 5 35m 6
28m 7 22m 8 16m
117
Advanced Optimization Techniques
Patching for High Frequency Correction
Middle Throw 1 61m 2 55m 3 48m 4 42m 5
35m 6 28m 7 22m 8 16m
118
Advanced Optimization Techniques
Patching for High Frequency Correction
Long Throw 1 61m 2 55m 3 48m 4 42m 5 35m 6
28m 7 22m 8 16m
119
Advanced Optimization Techniques
Patching for High Frequency Correction
Throw per speaker 1 61m 2 55m 3 48m 4 42m 5
35m 6 28m 7 22m 8 16m
120
Advanced Optimization Techniques
Environmental Conditions
8 M3D _at_ 60m. 20C, 50
121
Advanced Optimization Techniques
1 M3D _at_ 4m. 20C, 50
8 M3D _at_ 60m. 20C, 50AfterLow Freq Correcion
High Freq Correction
122
Advanced Optimization Techniques
123
Advanced Optimization Techniques
The End !!