SmaartLive Class Handout

SIA-Smaart ®
Schools & Applications Seminars
Presented by:
SIA Software Company, Inc
A LOUD Technologies, Inc. Company
www.SIAsoft.com
Copyright 2004
A LOUD Technologies Company
One Main Street, Whitinsville, MA 01588
All rights reserved

SmaartLive Fundamentals and Applications:
PowerPoint Notes
Presented by:
Copyright 2004 - SIA Software Company, Inc.
All rights reserved
SIA Smaart Schools & Applications Seminars
SmaartLive Fundamentals
PowerPoint: Class Notes
This PowerPoint presentation can be downloaded at:

http://www.siasoft.com/training/docs.shtml

All rights reserved
Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

One Main St Whitinsville, MA 01588 508.234.9877 ph. 508.234.7302 fax www.siasoft.com
1
SIA Contact Info

Jamie Anderson
SIA Product Manager
Jamie@SIAsoft.com
Calvert Dayton
SIA Development Manager
Calvert@SIAsoft.com
Barbara Stolakis
SIA Office Manager
Barb@SIAsoft.com
Other SIA E-mail Addresses:

SIA@SIAsoft.com
Support@SIAsoft.com
SmaartSchool@SIAsoft.com
Training@SIAsoft.com

One Main Street
Whitinsville, MA 01588
508.234.9877
508.234.7302
www.siasoft.com

2
FFT’s
Fast Fourier Transforms

3
Transforms
A transform converts data from one domain/view
to another: Time Domain to Frequency Domain
– Same data
• Is reversible via Inverse Transform
– Unlike a conventional RTA using a bank of analog filters,
FFT’s yield complex data: Magnitude and Phase data
Time Domain Frequency Domain
Waveform Spectrum*
Amp vs Time Amp vs Freq
(*Fractional Octave Banded View)
What do you get if you transform

a transfer function?
Inverse Fourier Transform (IFT) of Frequency
Response produces Impulse Response
Transfer Function To Impulse Response
Frequency Domain to Time Domain
*So . . . If Frequency Response can be measured with music . . .

so can Impulse Response*

4
FFT Resolution
• Reciprocal Bandwidth: FR=1/TC
Frequency Resolution = 1/Time Constant
– Larger Time Window:
• Higher Resolution
• Slower (Longer time window and more data to crunch)
– Smaller Time Window:
• Lower Resolution
• Faster
• Time Constant = Sample Rate x FFT Length
* Decimation – Varying SR & FFT to get constant res.*
The Only Math You Need Today:
T=1/ƒ
ƒ = 100 Hz ƒ = 250 Hz
T= 10 ms & T= 4 ms
ƒ=1/T T= .1 ms
ƒ = 10 kHz
T= .1 ms
ƒ = 10 kHz
ƒ = 20 Hz
T= 1 ms T= 50 ms
ƒ = 1 kHz
ƒ = 500 Hz T= .5 ms
T= 2 ms ƒ = 2 kHz

5
FFT Parameters:
Time Constant (TC) vs. Frequency Resolution (FR)
Linear Frequency Scale TC = FFT/SR

FR = 1/TC
FFT Parameters:
Time Constant (TC) vs. Frequency Resolution (FR)
Log Frequency Scale

6
Linear vs. Log Banding

FFT’s yield linear data
• Constant bandwidth instead of constant Q
• FFT data must be “banded” to yield fractional-octave data.
Pink Noise (equal energy per octave) shown w/ linear and log banding.
Linear banding has an increasing number of Fractional–octave (log) banding has

bands per octave as frequency increases, an equal number of bands per octave,
resulting in less energy per band in the HF. resulting in equal energy per band.
FPPO
24 Fixed Points Per Octave
(Only available in Transfer Function Mode)
TC = 3 ms
TC = 683 ms
FPPO mode utilizes multiple FFT’s of varying TC to

produce data that has a constant 24th oct. resolution.

7
Transfer Function
(Frequency Response)

8
Transfer Function
System
Input Signal Output Signal
Measurement
Channel (RTA)
Transfer
Function
Reference
Channel (RTA)
Transfer Function
System
Measurement
Channel (RTA)
Transfer
Function
Reference
Channel (RTA)

9
Transfer Function
System
Measurement
Channel (RTA)
Transfer
Function
Reference
Channel (RTA)
Dual-Channel Measurement Issues:

System
Delay = x ms
Ref Signal
• Propagation Time
• Linearity - Does response change with level?
• Noise
Averaging
Coherence

10
Coherence

11
Coherence
How stable/consistent is your data?
Coherence indicates the linearity/quality of each data
point in your transfer function measurement.
Given as a value between 0 to 1 (0% - 100%)

100% = Highest Coherence = great data
0% = Lowest Coherence = bogus data
Coherence Scale
100% (Top of Plot)
0% (Middle of Plot)
Coherence
Three causes of bad coherence:
1. Bad measurement
• Check measurement delay
• Check measurement signals
Look for broad ranges
• Check measurement set-up of bad coherence.
• Check equipment
Particularly in HF if
Smaart’s delay is set wrong.

12
Coherence
1. Bad measurement
2. Poor Signal to Noise Ratio
• Turn up measurement level If due to external noise,
coherence should improve
• Turn down “noise” with measurement SPL.
At what SPL should I measure?

How loud is loud enough?
(For accurate measurements)
Slowly turn up your measurement signal level . . .

When the coherence trace no longer improves, you’re there!
Coherence
1. Bad measurement
2. Poor Signal to Noise Ratio
3. Poor Direct to Reverb Ratio “Real World”
Coherence.
• Move mic closer to source
• Move source closer to mic
• Damp reverberance
It is common to get a
bad Coh “spike” where
you see a cancellation,

13
How Smaart Works

14
How Smaart Works
Input System Output
FFT Spectrum Views
RTA
FFT
Spectrograph
Wave Spectrum
SPL History
How Smaart Works
Input System Output
FFT
=
FFT Transfer Function
(Frequency Resp.)
Wave Spectrum

15
How Smaart Works
Input System Output
FFT
IFT
=
FFT Transfer Function Impulse Resp.
(Frequency Resp.)
Wave Spectrum
Basic Measurement Set-up
Loudspeaker
Source EQ / Processor Amplifier & Room
Microphone
Computer Mixer:
w/ Stereo Signal Selector
line-level input & Pre-amps

16
Mode by Mode:
Configuration Notes

17
Class Notes
Getting Started
•Point Smaart at your stereo input device
–Set as Wave Input in Options:Devices
–Set bit depth (Use 16 bit as default unless you know otherwise)
–Use [Alt + ”V”] to display the Windows Record In Panel
•Hit “Start” to begin processing inputs
•Verify that signals are getting to Smaart
–Ref. on Right (Blue) channel
–Meas. on Left (Green) channel
•Optimum signal input level is ~ -12dB
–Right where meters turn yellow
–Leaves enough headroom for dynamic signals
Class Notes Spectrum Mode

FFT Parameter “Set and Forgets”
Three goals for an RTA
1. Resolution
2. Responsiveness
3. Correlation to human hearing
FFT Size
• Larger FFT’s provide higher freq resolution
• Smaller FFT’s have faster response and easier to process
• Find the right trade off of resolution v. response (try 16k)
Sample Rate (SR)
• SR determines highest measurable freq = ½ SR = Nyquist
• SR determines time resolution = 1 sample = .02 ms at 48k
• Use highest SR for your default setting (48k)
Use Fractional-Octave Banding (Oct, 1/3, 1/6, 1/12, 1/24)
Fractional octave banding provides best correlation to human hearing

18
Class Notes
Spectrum Mode: SPL Calibration

1. View 1/3 octave scale
2. Use calibrator to generate tone of known SPL at mic.
3. Double-click anywhere in the RTA plot area
4. Type calibrator SPL into “Set this value to ___ dB”
Remember:
Smaart’s SPL See your Quick
calibration is dependent Reference Card for
upon mic sensitivity and alternative methods
input pre-amp gain. of SPL calibration
IFYOU CHANGE
EITHER YOU MUST
RECALIBRATE
Class Notes
Transfer Function Mode
Recommended Settings
Input Meters:
Optimum input level is @ -12
Where the meter turns yellow
Coherence Threshold:
10% - 15%
Just enough to remove the truly bogus data
Averages:
Acoustic Measurements: 64(+)
Electronic Measurements: 8 -16
More averages = better s/n & trace stability
Magnitude Threshold:
16 bit Input Device: 35%
24 bit Input Device: 55%
FFT Parameters:
Acoustic Measurements: FFT = FPPO
Electronic Measurements: FFT = 16k or 32k
Remember to set your delay!

19
Class Notes
Impulse Response
Three ways to improve the dynamic range of your

impulse response measurement:
1. Measure louder
Improve true s/n
2. Increase Averages
Each doubling of Avgs gives 3 dB better s/n
3. Increase TC/FFT Size

20
System Alignment Notes

21
Tools in Order of Use

• Acoustic Design / Treatment
1 • Equipment Choice / Maintenance
• System Design - “Design to align”
• Level
2
• Delay
3 • And lastly . . . EQ
System Engineering
Key Concepts:
• Systems interact most where they are equal
level.
– Phase/Time determines how they will interact.
• Solve the problem at the source.
• Use the right tool.

22
Comb Filters
&
Sine Wave Addition

23
Addition of Sine Waves

of Same Frequency and Equal Amplitude
²ø = 0°
Complete Addition
²ø = 90°
Partial Addition
²ø = 120°
No Addition
²ø = 180°
Complete Cancellation
²ø = 240°
No Addition
Period and Frequency

T = 4 ms
ƒ = 1/4 ms = 250 Hz
O ms 1 ms 2 ms 3 ms 4 ms
T = 3 ms
ƒ = 1/3 ms = 333 Hz
T = 2 ms
ƒ = 1/2 ms = 500 Hz
T = 1.5 ms
ƒ = 1/1.5 ms = 666 Hz
T = 1 ms
ƒ = 1/1 ms = 1000 Hz
T = .5 ms
ƒ = 1/.5 ms = 2000 Hz

24
Phase Shift vs. Frequency: Group Delay = 1 ms

1 ms
ƒ = 250 Hz
²ø = 90°
ƒ = 333 Hz
²ø = 120°
ƒ = 500 Hz
²ø = 180°
ƒ = 666 Hz
²ø = 240°
ƒ = 1000 Hz
²ø = 360°
ƒ = 2000 Hz
²ø = 720°
The Ground Bounce
Reflection arrives ~ 4 ms
4 ms
Comb filter frequency = 1/4 ms = 250 Hz

25
Reflection Arrival vs. Comb Freq.
1 ms 1 kHz
2 ms 500 Hz
4 ms 250 Hz
10 ms 100 Hz
The Ground Bounce

Two Solutions
Block the reflection
Remember:
Baffle must be large enough
to be effective above 100 Hz.
Think 5’x 5’ (1.5 m x 1.5 m)
Ground-plane measurement

26
Reading Delay in the Phase Trace

27
Smaart Measurement Delay

Time
System
Input Delay Measurement
Signal
Smaart Reference
Delay Signal
Phase Trace Angle Shows Delay
Time Phase Trace Angle
Meas
System
Smaart
Ref
Measurement Signal Lags Reference Signal

28
Meas
System
Smaart
Ref
Measurement Signal Aligned to Reference
Meas
System
Smaart
Ref
Measurement Signal Leads Reference Signal

29
Notes

30
SIA-Smaart ® Application Notes
Getting Started with SmaartLive:

Basic Measurement Setup and Procedures
Presented by:
All rights reserved
Getting Started with SmaartLive:
Basic Measurement Setup
and Procedures
Paul D. Henderson
This document serves as a starting point for learning to use SIA SmaartLive® for basic measurements
of audio systems and components. Here, we will discuss the capabilities of SmaartLive and the basic
measurement hardware necessary to perform successful measurements. A series of tutorial examples
will be presented, which will serve as a hands-on introduction to making measurements with the system.
I. The Primary Measurement Functions of SmaartLive

Fundamentally, SmaartLive is a software-based dual-channel audio analyzer, capable of performing a
large number of measurement tasks required by the audio professional. While SmaartLive is not intended
to replace critical listening and human experience, the intelligent application of the measurement platform
to the task of configuring, troubleshooting, and optimizing systems provides the user with significant
advantages. Table 1 outlines the basic measurement modes of SmaartLive, with a synopsis of some of
the common uses and capabilities for each.
SmaartLive Mode Primary Capabilities Applications

Spectrum • Real-time spectrum analysis • Live source spectrum monitoring
• Narrowband and fractional-octave display • SPL monitoring for live performance
• Calibration to real-world sound pressure • Noise level analysis
level with SPL metering • Feedback detection
• Running SPL log and spectrograph
functions
Transfer function • Real-time transfer function analysis • Transfer function measurements of loudspeakers,
• Configurable magnitude and phase equalizers, sound systems
display • Real-time optimization of systems (incl.
• Narrowband and Fixed-Point-per-Octave equalizers, crossovers, delays, etc.)
analysis
• Real-time coherence display
Impulse • Impulse response measurement • Measurement of sound system / room impulse
• Linear, log, and ETC display response
• Automatic estimates of propagation delay • Configuration of loudspeaker delays, etc.
times
Table 1: Basic measurement modes in SmaartLive.
In addition to the capabilities in Table 1, SmaartLive contains an internal signal generator that simplifies
the measurement process by creating the appropriate excitation signals for each measurement. This
eliminates the need for an external device dedicated to producing measurement signals for use with
SmaartLive. SmaartLive also includes significant capabilities for controlling external devices, such as
loudspeaker processors, equalizers, etc. This useful functionality will not, however, be discussed in
this document.
SIA Software Company, Inc. is not responsible for damage

to your equipment resulting from improper use of this
product. Be sure that you understand and observe the proper
input and output levels, impedances and wiring conventions
of all system components before attempting the
measurements described in this document.
Getting Started with SmaartLive: Basic Measurement Setup and Procedures Page 1
II. Components of a Basic SmaartLive Setup
To use SmaartLive effectively, you should have on hand a basic set of measurement equipment. The list
below outlines the fundamental components of a SmaartLive measurement setup. Details regarding the
connection of your equipment for specific measurement tasks will be presented in later sections.
Component Guidelines
Measurement If you intend to measure anything beyond simple electronic devices, you will need
Microphone a measurement microphone to acquire acoustical signals. The basic job of the
microphone is to convert acoustical pressure at a point into a voltage as accurately
as possible, so your microphone should be omnidirectional with a flat frequency
response. Most commercially available measurement microphones are based on
an electret condenser design, which will require some form of power, either
through phantom power from your preamplifier or by an internal battery. You
may also desire a microphone calibrator to accurately perform calibrated sound
pressure measurements.
Microphone To interface with your microphone, you will typically need some form of
Preamplifier microphone preamplifier. The preamplifier should possess a low noise floor
and sufficient gain for reasonable use. For most measurement microphones,
the preamplifier should include phantom power to power the microphone.
Note that many users may prefer to use a small mixer, routing device, or even a
front-of-house console in place of a dedicated preamplifier. In these cases, make
sure to disable all channel processing before use (equalization, dynamics
processing, etc.).
Sound Card For full functionality, SmaartLive requires a compatible sound card with at least
two independent line-level input channels (usually in the form of a single stereo
input) and a line-level output. Some notebook computers with built-in sound
capabilities only offer a single channel (mono) input, so make certain your system
meets this criterion. Without a stereo input, you will be unable to utilize the
transfer function and impulse measurement capabilities.
Note that external hardware solutions are available, some of which combine high-
quality A/D and D/A converters with built-in microphone preamplifiers. These
may prove to be maximally convenient for field use, are readily available with
USB, PCMCIA, and Firewire interfaces. Regardless of the input device you
select, it must use a Windows®-compatible Wave audio device driver. Other
device driver types, including ASIO, are note currently supported by SmaartLive.
Computer with The computer should adhere to at least the minimum requirements for running
SIA-SmaartLive™ SmaartLive, which are available in the SmaartLive User Guide or by accessing
the SIA Software website at http://www.siasoft.com. For portable field operation,
a notebook computer is most convenient.
Cabling and You should have on hand the appropriate cables for connecting your measurement
Interconnections system and interfacing with the equipment that you are measuring. Use only
professional-quality cables, avoiding inferior adapters and consumer-grade
interconnections. If your computer sound card uses 3-conductor 1/8-inch stereo
phone connectors for interface, you may obtain breakout cables that allow you to
convert this interface to 1/4-inch phone or XLR connectors.
III. Getting Signals into SmaartLive
Now that you have assembled the equipment necessary to operate
SmaartLive, we will discuss the process of setting up the system
to recognize your hardware and adjusting the signal levels
through the system. Make sure your sound card is online (if
external, connect it) and start SmaartLive. If you are using an
external audio device, do not disconnect it while SmaartLive is
running. To select the proper sound card for use by the system,
click Options? Devices on the menu, or press Alt+A. The
window in Figure 1 will appear. Select your sound card input
device in the Wave In drop-down box, and do the same for the
output device in the Wave Out box. If your device supports input Figure 1: Select your sound card as the input
or output resolution higher than 16 bits, select the appropriate and output device.
values in the Bits Per Sample boxes. Now, you may connect
devices to the outputs and inputs of your soundcard, and the
signals will be correctly handled by SmaartLive.
Many soundcards use internal mixing circuitry to mix many

audio streams together to the master output, or to select and/or
mix the signals together that will be seen by the sound card
inputs. You may need to configure these options for your card
to enable the line-level input and also to enable the wave
output. These options may be set by launching the Windows®
mixer application, seen in Figure 2. You may launch the mixer
by pressing Alt+V on the keyboard in SmaartLive. See the
SmaartLive User Guide for more detailed information.
(a) (b)
Figure 2: Configure the input (a) and output

A note about levels: the signal generator level is controlled by (b) mixers for your device if necessary.
the level setting spinner on the generator, which defaults to a
low level to prevent equipment damage at first use. This, in combination with output levels and mixer
settings for your device, varies the signal level seen at the sound card output. For most measurements
where the signal generator is required, you will desire to adjust the overall generator level to a high
enough level to avoid noise in the measurement, but not so high a level as to cause damage or discomfort.
The input levels, however, must be carefully adjusted to

eliminate signal clipping on the sound card A/D converters while
maintaining a high enough signal level to minimize extraneous
û ü û
noise. At any time SmaartLive is running, the input level meters
in Figure 3 are active, which indicate the peak input levels seen
by the A/D converters in the sound card. If the levels are too
low (Figure 3a), noise from the sound card and other analog
devices may compromise the measurement. If the levels are two (a) (b) (c)
high, the clipping indicators will light (Figure 3c), and you must
reduce the input level to perform meaningful measurements. Figure 3: Make sure input levels are neither
too low (a) nor too high (c). Input levels
For most measurements, maintain a nominal input level between should usually be around -12 to -6 dB for basic
-12 and -6 dB. measurements (b).
IV. Measurement Examples
Now, we will explore some examples of basic measurements using SmaartLive. The following pages will
introduce you to the fundamentals of making Spectrum (RTA), Transfer Function, and Impulse Response
measurements. The examples presented here are arranged in order of complexity; the later examples build
on concepts presented in preceding sections. We recommend that you proceed through these exercises
in order. Each example will present the basic hardware configuration, as well as sample measurement
results. Keep in mind that your data may appear different than the data presented here; the sample results
are simply representative of what you might see.
Example Application 1
SmaartLive as a Real-Time Spectrum Analyzer (RTA)
The most basic functionality of SmaartLive lies in its Spectrum mode, which enables two channels of
real-time spectrum analysis. In this mode, SmaartLive contains functions similar to a hardware RTA
(real-time analyzer), where the incoming signals are decomposed into frequency components and
displayed dynamically. By default, SmaartLive displays the two channels as a real-time bar graph of
energy vs. frequency, with each bar representing a band of energy 1/12th-octave wide, although many
other displays are possible.
Connecting the System

To make use of this mode, connect any compatible audio signal to at least one input of your sound card.
In RTA mode, you may monitor two channels simultaneously, so you could connect a measurement
microphone to one channel and the output of a mixing console to the other, among other possibilities.
Measurement
Microphone
Microphone Line
Preamplifier Input
L
Computer with
R SmaartLiveTM
Mixing Console
Figure 4: Example hardware configuration for RTA measurement.
Figure 4 shows the hardware configuration for this example. Any line-level
signal may be used as an input source, although it is somewhat educational to perform
this introductory measurement with a live microphone connected through a preamplifier.
Now, launch SmaartLive, which should default into Spectrum mode. At any time, you Measurement
may change the current measurement mode by clicking one of the mode buttons. Press mode buttons
the ON button to begin the real-time measurement.
You should now see a live banded spectrum of
the input signals as in Figure 5 (both channels
are visible by default). If you are using a live
microphone, the spectrum display will respond
to any noise in the room, and whistling near the
microphone will drive the bands noticeably upward
near the frequency of your whistle.
You may experiment with

different views of the incoming
data by selecting a different
data scale with the Scale spinner;
data may be presented in 1/1,
1/3, 1/6, 1/12, and 1/24-octave Figure 5: Default live RTA spectrum display (1/12th-octave).
bands or as a narrowband power spectrum on either
a logarithmic or linear frequency axis. Also configurable are the basic Fast Fourier Transform (FFT)
parameters, averaging, and weighting. Averaging may be used to vary the time behavior of the spectrum
display, allowing you to look at either the instantaneous behavior of the signal or its long-term spectral
content. Averaging may be configured with various options, including Slow, Fast, and Exp, which
represent exponentially decaying behavior, Inf, which maintains an evenly-weighted average of all data
since initialization, and numeric values between 1 and 128, which evenly weight the last n measurements.
You may “reseed” the averaging buffers at any time by pressing V on the keyboard.
The spectrum mode contains many other useful features, including absolute calibration, integrated SPL
metering and logging, and real-time spectrogram functions (Figure 6). Note that these functions will
display data for the active input channel only. The active channel may be assigned by clicking on the
associated input level meter. Please see the included documentation for more information on the use
of these more advanced functions.
SmaartLive’s Spectrum mode displays are very useful for identifying feedback frequencies, looking at
room noise, studying the spectral content of musical material, and has many other uses. Historically,
many have employed RTA methods for measuring the frequency response of a system and performing
equalization; we do not recommend using this technique, as the transfer function measurement capability
of SmaartLive is a far more useful and accurate tool for this task. The Spectrum mode is inherently unable
to distinguish direct sound from reflective energy and to discriminate between the excitation signal and
uncorrelated noise, which limits its usefulness in sound system response optimization tasks.
(a) (b)
(c)
Figure 6: Sample advanced Spectrum-mode measurements;

(a) Spectrograph, (b) SPL time history, (c) Real-time calibrated SPL.
Measuring an Analog Equalizer
In this example, we will introduce the use of SmaartLive’s real-time Transfer Function measurement
capability to measure the frequency response of an analog equalizer. To perform this exercise, you will
need an equalizer, crossover, or some other filtering signal processor. An analog device is best for this
example; digital devices include some throughput delay from input to output that must be found and
compensated for, using SmaartLive’s internal signal delay (we will discuss this in subsequent examples).
Equalizer
Out In
Line Line
Input Output
L L
Computer with
R SmaartLiveTM R -or-
Figure 7: Hardware configuration for measuring an equalizer.

Connect the system as shown in Figure 7, with the sound card output driving both the input of the
equalizer and the right input of the sound card (called the reference input). The output of the equalizer
is brought into the computer through the left input (called the measurement input). The internal signal
generator in SmaartLive will be used to excite the equalizer and measure its response. By using this
connection scheme, SmaartLive compares the exact representation of the generator output to the returning
signal from the equalizer, effectively canceling any imperfections introduced by the sound card.
Adjusting Signal Levels

Start the SmaartLive generator by clicking the GEN (Generator On) button; the system
will now begin generating random pink noise that will appear on the sound card output. The generator
defaults to a low level (-36 dB), so use the generator level spinner to increase the output to a reasonable
level. The level should be high enough that the dynamic range of the system and equalizer is adequately
utilized; try for a generator level of around -6 dB, as a rule of thumb target, to optimize the performance
of the internal generator.
If the analyzer is not running, click the ON button to start it. Now, adjust the input controls to bring the
input signal to a reasonable level (as in Figure 3). If the equalizer is bypassed or its controls set to 0 dB,
the transfer function display should be an approximately flat line. If not, it is likely that there is some gain
error in the system or the device itself contains some gain or attenuation factor. You may
adjust the measurement channel gain independently to correct for this, or use the dB +/-
spinner to adjust the “zero” level of the displayed transfer function.
Performing the Transfer Function Measurement
SmaartLive’s Transfer Function mode measures a system’s frequency response by comparing its input
signal to its output signal. This measurement shows the difference between those two signals in both
magnitude and phase and represents the processing behavior of the system as a function of frequency.
Gain and loss show up as deviation from the center 0 dB line on the magnitude trace.
By adjusting filter settings on the equalizer, you should be able to see the changes being made in
the frequency domain on the SmaartLive display. If attenuation (or “cut”) on the equalizer shows up as a
peak on the display, it is likely that you have inadvertently reversed the input signals. You can either
physically swap the input cables or press the Swap button to reverse the signals and obtain the proper
display.
SmaartLive defaults to the FPPO (Fixed-Point per Octave) transfer function mode, which
provides you with transfer function measurement points distributed equally on a logarithmic frequency
scale by varying the FFT length at different frequencies. You may wish to experiment with different
fixed-width FFT parameters, sampling rates, and excitation signals to see the effect of the various
parameters. You may also press the Phase button to see the transfer function phase (in addition to
magnitude) as a function of frequency. Figure 8 shows an example measurement of a single parametric
equalizer filter, including the phase plot.
Figure 8: Sample measurement of an analog parametric equalizer filter.
Note that, for this measurement, we have not included any compensation for delay through the equalizer,
since the delay through almost any analog equalizer will be insignificant relative to the length of the FFTs
used in the calculations. When measuring a digital device, loudspeaker, or almost any electroacoustic
system, we must first measure the propagation delay time and use SmaartLive’s internal alignment delay
to compensate for any delay in the external system. This will be covered in Example 3, “Measuring
a Loudspeaker”.
Measuring a Loudspeaker
In this example, we will use both the Impulse and Transfer Function modes in SmaartLive to measure
the behavior of a loudspeaker in a room. To perform this exercise, you will need a loudspeaker and power
amplifier (or powered loudspeaker) in addition to your SmaartLive measurement system.
Measurement
Microphone
Loudspeaker
Microphone Line Line Power

Preamplifier Input Output Amplifier
L L
Computer with
R SmaartLive TM R -or-
Figure 9: Hardware configuration for measuring a loudspeaker.

Connect the system as shown in Figure 9 above. The output of the sound card drives the power amplifier
and loudspeaker as well as the right input (the reference signal) of the computer as in the preceding
example. For this measurement, however, a measurement microphone is used as the measurement signal,
which is brought into the left sound card input via a microphone preamplifier.
You will want to position the measurement microphone at a nominal distance from the loudspeaker, for
example, 1m. The farther the microphone is from the loudspeaker, the more difficult it is to separate the
direct sound of the loudspeaker from the influence of reflections in the room.
Adjusting Signal Levels
We must now adjust the signal levels to obtain a reasonable gain structure through the system. Enter
SmaartLive’s Spectrum mode by clicking the Spectrum button and turn the analyzer ON. As before, turn
on the internal signal generator and adjust its level until the sound level from the loudspeaker is
appreciably higher than the ambient noise in the room. You may wish to start with a low power amplifier
gain and then slowly increase the level to prevent unexpectedly loud signals from reaching your ears.
Next, adjust the input levels on the sound card and the microphone preamplifier gain to achieve a proper
input level, following the guidelines in Figure 3. For best results, you will want to match the levels at the
reference and measurement inputs as closely as possible, so adjust the preamplifier gain to achieve this.
Performing and Interpreting an Impulse Response Measurement
Now, we will perform a basic impulse response measurement of the loudspeaker in the
room. Click the Impulse button to enter the Impulse mode; SmaartLive will automatically measure an
impulse response of the system and display it in the graph window. You may make another impulse
response measurement at any time by pressing the Start button. An example of what you might see is
shown in Figure 10.
The impulse response in Figure 10 is quite typical of
what you might see when measuring a loudspeaker
in any non-anechoic room (any room with reflective
surfaces). The impulse response view shows a time
history of energy arriving at the microphone, and is Direct Sound
very useful for understanding exactly what you are
measuring. Figure 10 shows the Log magnitude Room Reflections
view, which indicates the magnitude of the impulse Noise
response in dB. SmaartLive can also display the
impulse response in Linear units (which preserves
polarity information) and as an ETC (Energy-Time
Curve), which extracts the decay envelope from the
impulse response, displayed in dB. The large peak at
the beginning of the plot in Figure 10 indicates the
arrival of direct sound from the loudspeaker, which Figure 10: Impulse response of a small loudspeaker in a room.
in this case, is the component that interests us.
SmaartLive automatically detects the time and
magnitude of this peak, which, in this case, is arriving with a delay time of 2.29 ms. This delay is due to
the propagation time from the loudspeaker to the microphone through the air over distance of
approximately 2½ feet. We will use this concept in correcting for the propagation delay when we perform
a transfer function measurement. The other energy shown in impulse response is due to reflections in the
room and noise in the measurement. The noise floor can be seen to have a constant average level; the
accuracy of your measurements is dependent upon an adequate signal-to-noise ratio between the direct
sound and this noise level.
Performing a Transfer Function Measurement of the Loudspeaker

Next, we will measure the frequency response function of the loudspeaker using the
Transfer Function mode in SmaartLive. Click the Transfer button to switch to the
transfer function mode, and make sure the analyzer is running. We now need to set
SmaartLive’s internal delay to compensate for the propagation time between the
loudspeaker and microphone.
Click the Auto Sm (Delay Auto-Locator Small) button to do this; SmaartLive will run an
impulse response measurement in the background and automatically measure the delay time. The “Delay
Found” dialog box appears with the measured delay time; click “Insert Delay” to accept the shown delay
time to compensate for the propagation delay during transfer function measurements. If the shown delay
time seems impossibly long, you likely have the reference and measurement inputs swapped; simply
swap the input cables and try again. This process may also be performed manually using the Impulse
mode; click the Set Delay to Peak button in the impulse response mode
to set the transfer function delay to the peak arrival time.
You may use the delay presets feature to store several

delay times for quick recall using the keyboard. This is useful if you are
switching between microphones in different locations or between a
measurement microphone and an equalizer, for example. To assign the
current value to a present, click the Delay label above the delay readout
field and then click on one of the five buttons labeled F6-F10 on the
delay tab of the options dialog box. You may then recall this delay at
any time by pressing the corresponding function key on the keyboard.
For this example click on the button marked F6 to
store the current delay into the first preset. Press
OK to exit the dialog box.With the analyzer running
in transfer function mode, you should now see the
frequency response of the loudspeaker (energy vs.
frequency) displayed in the window as shown in the
example measurement in Figure 11. You may find
that the display is somewhat
erratic; increasing the number of
averages using the Avg spinner will stabilize the
transfer function trace. Set the number of averages
16 or higher to see improved behavior.
Saving a Measurement Trace

Figure 11: Sample transfer function measurement of a small
using the Reference Registers loudspeaker.
SmaartLive’s Reference Registers are used to capture and store “snapshots” of the active live trace. The
Reference Registers are represented by five groups of small solid-color buttons, labeled A, B, C, D and E,
located below the plot area.
Click the button for register A1 (the first register button in the A group). This “activates”
the register even if the button was already depressed. Click the Capt (Reference Capture)
button below the plot area to sample and display the current trace as an overlay on the plot.
Click the A button to remove the captured trace from the display. The sampled trace, called a Reference
Trace will remain stored in the register until you erase it or
capture another trace to the same register.
To permanently save a Reference Trace to a file on disk,

called a Reference File, click the Info button to the right of the
capture button. This opens the Reference Trace Information
dialog box. This dialog box has six tabbed “pages.” Click on
the tab labeled A in the upper portion of the dialog box to bring
that page to the front. Select the register containing the
Reference Trace you just captured (by clicking the first of the
four solid-color register buttons on the left) and click the Save
button. This opens a Windows Save file dialog box prompting
you to select a file name ending with the (*.ref) extension.
Reference Files may be recalled later and displayed as traces by selecting a register in this same dialog
box and pressing the Load button. You can also save and reload the contents of all 40 reference registers
as Reference Group (*.rgp) files using the Save All and Load All buttons on the General tab of this dialog
box. After saving the reference trace to a file, click the OK button to exit the Reference Trace Information
dialog box.
Note that when you capture a reference trace, the stored trace is initially displayed “in
front” of the live trace on the plot. The text color in the dB +/– spinner field to the right of the plot
changes to match the reference trace color and when cursor tracking is enabled, the mouse tracking
cursor follows the stored trace instead of the live trace. You can return the focus of the display to the
live transfer function trace by clicking anywhere on either input level meter with your mouse.
Equalizing a Loudspeaker
In this example, we will use the transfer function measurement from Example 3 to set an equalizer
to optimize the performance of the loudspeaker. To perform this exercise, you will need the equalizer
in addition the loudspeaker and power amplifier from the previous examples.
Measurement
Loudspeaker
Microphone
Line Line Power

Input Output Equalizer Amplifier
EQ L L
Computer with
mic R SmaartLive TM R -or-
Microphone
Preamplifier
Figure 12: Hardware configuration for equalizing a loudspeaker.

Connect the system as shown in Figure 12 above. When compared to the last example, this configuration
simply inserts the equalizer into the signal chain between the sound card output and the power amplifier.
A “return line” is also included, which simply receives the signal that is present after the equalizer,
allowing you to adjust the filters in real-time exactly as in Example 2. The switch shown on the left sound
card input is simply to show that you will wish to select between the measurement microphone signal and
the signal coming directly from the output of the equalizer. This may be accomplished in many ways,
including the use of a small mixer, a physical switch, or by simply re-patching the input cables.
Performing the Measurements and Setting the Equalizer

We will assume that you have performed the measurement of the loudspeaker as described in Example 3,
and that the transfer function trace from the measurement is stored in reference register A1 with the
propagation delay time stored into the first delay
preset (recallable by pressing F6 on the keyboard).
If this is not the case, perform the measurements
described in Example 3 before proceeding.
Now, reconfigure the inputs such that the left or

measurement input is taken from the output of the
equalizer. Make certain that the input and output
levels are appropriate and no clipping is present
on any device. If you are using an analog equalizer,
press F5 on the keyboard to reset the internal
delay to 0 ms. If you are using a digital equalizer
(and/or digital mixer), click the Auto Sm button
to measure and insert the appropriate delay time
Figure 13: Real-time inverted equalization filters (yellow)
into SmaartLive. matched to the measured loudspeaker response (orange).
To equalize the system, we will calibrate the equalizer to have a frequency response that is approximately
inverse to the response of the loudspeaker. In other words, the peaks in the loudspeaker response will
correspond to equivalent nulls in the equalizer response, flattening the frequency response curve. We
will use the stored reference trace containing the measured loudspeaker response as a template by which
to adjust the equalization filters.
Make sure that the reference trace containing the

loudspeaker response measurement is active on
the display (in this case, A1) by using the reference
trace control buttons. Use the dB +/- spinner to
center the curve vertically near 0 dB. Now, click
the Swap button to invert the measured frequency
response of the equalizer. Any cut filters will now
be displayed as boosts, and vice-versa, which
allows us to visually match the inverted equalizer
response to the loudspeaker transfer function to
optimize the equalizer settings. Now you may
adjust the equalizer to best fit the inverted equalizer
response to the measured loudspeaker transfer
function (Figure 13). After the equalizer settings
Figure 14: Resulting equalized loudspeaker response after
have been optimized, you may perform another applying the filters from Figure 12.
transfer function measurement of the entire system
to see the final result (Figure 14). To perform this measurement, select the measurement microphone as
the measurement input, and recall the appropriate delay preset by pressing F6 on the keyboard.
As a practical note, boost filters are best used sparingly when optimizing the frequency response of a
sound system. Excessive use of boost filters may destabilize a sound system by reducing gain-before-
feedback and/or headroom, as the nulls that you observe in measurements may not be present in all spatial
locations (due to comb filtering from reflections, room modes, loudspeaker interference, etc.). If you find
that you require a large number of boost filters or that any required filter is very narrow in bandwidth,
your problem may not be best resolved through equalization alone. Correction of acoustical conditions,
crossover settings, or loudspeaker arrangements may be necessary. Typically, electro-acoustical
phenomena that produce wide bandwidth (low-Q) peaks in frequency response are most effectively
addressed through equalization. Additionally, we recommend the use of parametric equalizers for
precision equalization, to enable the selection of proper bandwidth and center frequency for each filter.
© 2003 SIA Software Company, Inc., Whitinsville, MA USA

All rights reserved. www.siasoft.com
The Fundamentals of FFT-Based

Audio Measurements in SmaartLive
Presented by:
All rights reserved
The Fundamentals of FFT-Based
Audio Measurements in SmaartLive®
Paul D. Henderson
This article serves as summary of the Fast-Fourier Transform (FFT) analysis techniques implemented in
the SIA-SmaartLive® measurement platform. By reading through this document, you will receive an
understanding of the fundamental concepts in FFT-based measurements used throughout the SmaartLive
application, providing you with insights to better comprehend the measurement parameters, procedures,
and resulting data. As a prerequisite to this text, you should be familiar with the basic concepts presented
in the article “Getting Started with SmaartLive: Basic Measurement Setup and Procedures”.
Time Domain Sampling: Getting Signals into SmaartLive

Most of the acoustical and electrical signals that we may wish to measure are signals that are continuous,
that is, they have a defined value for every possible instant in time. Sound pressure and analog voltage
are two examples of continuous-time signals that we may wish to measure using our instrumentation.
However, in order for these continuous signals to be analyzed using a computer-based measurement
system such as SmaartLive, we must convert the signal into a stream of digital samples, with each sample
representing a numeric value that is proportional to the measured signal at a specific instant in time. This
process is called sampling: converting the continuous-time signal into a discrete-time signal (a process
handled by an analog-to-digital converter in the computer sound card).
The sampling process employed for SmaartLive measurements (and for most other purposes in digital
audio) creates digital signal data spaced on an even interval of time. The number of samples per second
is the familiar sampling rate (or sampling frequency), referred to here as SR in units of Hz. The sampling
rate directly affects the highest frequency that we may analyze in the
computer, conventionally called the Nyquist limit frequency (fmax), f max = SR
which is exactly equal to one-half the sampling rate. For measurements 2
fmax = Nyquist limit frequency (Hz)
on electroacoustical signals and systems, we are most interested in SR = sampling rate (samples/sec)
signals that lie in the frequency band from approximately 20 Hz to 20
kHz (the range of human hearing). Therefore, for most measurements,
you will wish to choose the highest sampling rate compatible between
your sound card and SmaartLive, typically either 48 kHz or 44.1 kHz,
which will provide a measurement bandwidth of at least 20kHz. T
A parameter inversely related to the sampling rate is the sampling

period (T) which refers to the length of time (in seconds) between
samples. Our measurements will have a time resolution equal to T,
meaning that we cannot discern any details from the time signal that
have a duration less than this value. For example, if we use SmaartLive SR = 1
to measure signal delays by finding peaks in the impulse response, we T
cannot measure a time delay difference of less than T. SR = sampling rate (samples/sec)
T = sampling period (sec/sample)
In addition to the choice of an optimal sampling rate, we must also

consider the word length of the analog-to-digital converter when configuring the measurement system.
For every sample, the analog-to-digital converter must assign a defined digital bit pattern to represent the
The Fundamentals of FFT-Based Audio Measurements in SmaartLive® Page 1

amplitude of the signal at that instant. In effect, the longer the word length, or bits per sample, the higher
the dynamic range of the measurement. In addition (and perhaps more importantly), increasing the word
length increases the amplitude resolution of the measurement, as the sampled amplitude steps are
distributed on a smaller interval. Inexpensive sound cards may offer word lengths of 16 bits, with
professional cards offering 24 bit input and output capability. As with the sampling rate, set the word
length in SmaartLive to the maximum values compatible with your hardware.
üQuick Reference
Sampling rate (SR) The number of samples per second (in Hz) used in the conversion process. Sets the
maximum frequency that may be analyzed (Nyquist=SR/2). Set to 48kHz or 44.1kHz
in SmaartLive for most measurements.
Sampling period (T) The time interval between samples, equal to 1/SR seconds. Time-domain details of
duration less than T (sec) will be masked (delays, reflections, etc.).
Word length Number of bits used by the analog-to-digital and digital-to-analog converters in the
sound card. Always use the maximum values compatible with your hardware.
FFT Analysis: Viewing Frequency Domain Information

Optional Note: Fourier Theory
While performing measurements in the time domain may
be interesting and useful in many cases, we require more The 19th century French mathematician Jean Baptiste
information for most audio measurements: spectral Joseph Fourier proposed a concept allowing us to
information, or knowledge about the frequency content express any time signal as a function of fundamental
frequencies. Fourier theory states that any complex
and behavior of the audio signals and of complete
time signal, be it noise, speech, music, etc., is
systems. Fortunately, there exists a defined technique composed of a combination of sinusoidal waves of
for converting, or transforming data from the time varying frequency, amplitude, and phase. We can use
domain into the frequency domain, where information this basic concept as a transform, or a mathematical
exists about the spectral content of signals. The Fourier method for moving signals between the time and
frequency domains.
Transform allows us to convert a time signal to the
complex frequency domain, meaning the spectral data To convert a continuous signal x(t) to its frequency
contains information about both the amplitude and phase domain counterpart X(j? ), we may use the forward
of the sinusoidal components that make up the signal Fourier Transform:
+∞
(see sidebar). In addition, the Fourier Transform
∫ x(t )e
− jωt
provides us with an inverse transform, which allows us X ( jω ) = dt
to convert the complex frequency-domain signal data −∞
We may reverse this operation with no loss of
back into the time-domain without losing information. information using the Inverse Fourier Transform:
Therefore, the both the time- and frequency-domain data +∞
1
∫ X ( jω ) e
jωt
are equivalent: the two domains simply provide a x(t ) = dω
different view of the same signal (see Figure 1). 2π −∞
It should be noted that, in the most formal sense, the
SmaartLive uses a digital implementation of the Fourier Fourier Transform requires the complete time history
Transform called the Fast Fourier Transform, or FFT, of a signal (for all time: an infinite-length view) and an
infinite number of sinusoidal frequency components to
which is simply a computationally-efficient method for fully describe a signal. This is, obviously, of no
computing the Fourier Transform on digital signals. The practical use for measurement, since we can only
FFT works on finite-length blocks of sampled data observe the signal for a finite amount of time. To
(called FFT frames). We will use the notation NFFT for implement the Fourier Transform computationally, we
the length of the FFT data frame in samples (or, must utilize time windowing to limit our view of the
signal to a finite frame of time. We may utilize the
equivalently points). From this value, we can easily Discrete Fourier Transform (DFT), which operates on
compute the length of the FFT frame in units of time sampled-data signals, or the Fast Fourier Transform
(FFT), which computationally accelerates the DFT.

using the sampling rate. We will call this value the time constant TC,
representing the length of time that each FFT frame observes the TC = N FFT
continuous input signal. Higher the values of TC provide a longer time SR
“window” in which we observe the signal, but also increase the time TC = FFT time constant (sec)
NFFT = FFT size (samples)
between subsequent updates of the FFT spectrum and the amount of SR = sampling rate (samples/sec)
numeric data that must be processed.
Amplitude vs. Time Magnitude vs. Frequency

Voice Signal
FFT
Inverse FFT
Time Waveform Signal Spectrum

Time Frequency
Domain Domain
Amplitude vs. Time Magnitude and Phase vs. Frequency
System Response
FFT
Inverse FFT
Impulse Response Transfer Function
Figure 1: The Fourier Transform: moving signals between the time and frequency domains. Upper example shows the
conversion between an impulse response and a transfer function; lower example is the conversion between a
time signal (voice sample) and its spectrum.
The FFT generates frequency domain data that is linearly-spaced as a

function of frequency, unlike our hearing, which primarily perceives FR = 1
TC
frequency in a logarithmic manner. Conveniently, the time constant FR = Frequency resolution (Hz)
TC is inversely proportional to the frequency resolution of the spectral TC = FFT time constant (sec)
data, referred to here as FR. The FFT spectrum will include complex
frequency data (magnitude and phase) spaced on even intervals of FR
extending from 0 Hz (DC) to the Nyquist frequency. With this concept, it is easy to see how resolution is
inversely related between the time and frequency domains. Using longer FFT sizes provides higher-

resolution spectral data but more “sluggish” time response, while shorter FFT sizes provide lower spectral
resolution but faster time response. Figure 2 graphically demonstrates the effect of changing the FFT
parameters on resolution in the frequency domain. When distributing the linearly-spaced FFT data on a
logarithmic axis, it can be easily seen that short FFT’s may provide inadequate low-frequency resolution,
while long FFT’s may provide excessive high-frequency resolution.
Frequency (Hz) Frequency (Hz)

(Linear Scale) (Log Scale)
Figure 2: The effect of FFT parameters on frequency resolution. Note that the FFT spectrum data is equally-spaced
on a linear frequency scale but exponentially-distributed on a logarithmic frequency scale. This may yield
inadequate low-frequency resolution for short FFT’s and excessively detailed high-frequency resolution for
long FFT’s.
Fortunately, we are not required to manually calculate the parameters that have been
discussed here. The FFT Parameters function in SmaartLive allows for the independent
selection of sample rate, FFT size, time constant, and frequency resolution. A change in
any one parameter immediately updates the other dependent parameters, allowing the user
to concentrate on the meaning of the values, not on their calculation. An indicator in the
main SmaartLive window shows the active settings for the live input signals.
üQuick Reference
FFT The Fast Fourier Transform, a method for moving digital signals between the time
and frequency domains.
FFT length (NFFT) The length of the FFT input data frame in samples.
Time constant (TC) The length of the FFT input data frame in seconds, equal to NFFT/SR. Indicates the
length of time that the FFT observes the signal in each data frame.
Frequency resolution The frequency resolution of the FFT spectral data, in Hz, equal to 1/TC. FFT data is
(FR) linearly spaced from 0Hz to the Nyquist limit on even intervals of FR.
Improving the Spectral View

We have shown that the frequency domain spectral data from an FFT is distributed on a constant,
linearly-spaced interval in frequency. If we evenly distribute the FFT data onto a graph, we will need to
employ a frequency axis that is equally spaced in frequency, that is, there is an equal interval per Hz. In
contrast, our hearing perceives musical pitch on (approximately) equal frequency ratios, yielding a
logarithmic frequency axis for equal-interval analysis (an equal interval per octave). Obviously, if we

distribute the linearly-spaced FFT data on a
logarithmic frequency axis, the apparent resolution Fractional-Octave Banding
along the frequency axis is not constant, possibly The Spectrum mode banding technique provides an FFT-
with insufficiently low resolution in the lower based approximation of analog Real-Time Analyzer (RTA)
octaves, and inappropriately high resolution in the filters, which utilize electronics to separate the incoming
upper octaves (as seen in Figure 2). We can spectrum into bands and then derive the RMS level of each
band over time. The banded views in SmaartLive provide
maximize the low-frequency resolution by an accumulation of the total energy in each logarithmically-
increasing the FFT size, however, this leads to a proportioned band (just like a hardware RTA), so signals
longer time constant and, consequentially, slower with constant power per octave (like pink noise) will result
time response. In addition, for transfer function in a flat banded spectrum (see figure below). Signals with
constant power per Hz (like white noise) will indicate a
measurements, the increased time window will also
rising (+3 dB/octave) banded spectrum. The banded
include additional room reflections in the spectrum view is very useful for identifying the generic
measurement, potentially obscuring details in the frequency content of input signals, identifying feedback
response curve. Obviously, we require additional frequencies, and investigating room noise levels. However,
methods for viewing the FFT data in a perceptually- system equalization and tuning procedures are best
performed using the transfer function measurement mode.
significant manner to better correlate the graphical
data with human hearing.
SmaartLive contains several methods for improving

the logarithmic-frequency view of FFT data, which
vary according to the analysis task at hand. For
spectrum-mode data (where we are simply interested
Narrowband Log Spectrum 1/3-octave Banded Spectrum
in the spectrum of the input signal), we may employ
frequency-domain banding, which distributes the Pink Noise Spectrum: Note the apparent downward slope seen
when viewed as a narrowband spectrum becomes a flat curve in the
data into equally-spaced fractional octave bands. log-proportioned banded view. In this case, the banded view more
The spacing and, consequentially, the width of each accurately reflects the perceived timbre of the signal.
band is selectable as either 1-, 1/3-, 1/6-, 1/12-, or
1/24-octave. Figure 3 compares the log-distributed FFT spectrum with the banded spectrum for an input
signal.
(a) (b) (c)
Figure 3: Spectrum-mode frequency banding; (a) original FFT spectrum data distributed onto a log-frequency axis,
(b) spectrum data displayed in 1-octave bands, (c) spectrum data displayed in 1/24-octave bands.
In addition to the banding method for handling spectrum-mode signals, the Fixed-Point per Octave
(FPPO) technique in SmaartLive is used for computing logarithmically-spaced Transfer Function
(frequency response) data. In effect, the technique utilizes a measurement time window that varies as a
function of frequency, utilizing a long time window at low frequencies (for narrow frequency resolution)
and a successively shorter time window at high frequencies. This method has two main effects: the
variable time window is well-correlated with the hearing perception mechanism defining the perceived

spectral quality of a loudspeaker operated within a room, and the variable analysis frame lengths provide
frequency-domain data that is of equal density per octave, in this case, 24 points per octave.
For most transfer function measurements in SmaartLive, especially those involving some acoustical path
(a loudspeaker or sound system measurement), the FPPO view provides the best representation of the
system response function. Beyond the inherent low-frequency advantages, the FPPO technique typically
provides a more easily readable trace at high frequencies, in contrast to the “fuzzy” character of standard
FFT data caused by excessive high frequency resolution (see Figure 4).
(a) (b)
Figure 4: Log-frequency transfer function analysis; (a) transfer function of a small loudspeaker using a 32k-point
standard FFT, (b) measurement of the same loudspeaker using the FPPO technique.
In SmaartLive’s transfer function mode, an additional option exists for smoothing of the transfer function
trace over a definable number of points. The smoothing function is, effectively, a moving average filter
that is applied to the transfer function data before it is displayed in order to minimize the presence of
jagged edges and discontinuities in the displayed data. You may select either 3-, 5-, 7-, or 9-point
smoothing depths, which define the number of FFT data points surrounding an individual value that are
averaged to derive the displayed value; higher numbers yield a more continuous visual curve. Figure 5
shows an example of smoothing applied to a transfer function measurement of a small loudspeaker.
Trace smoothing is available for both standard FFT sizes and FPPO curves.
(a) (b)
Figure 5: The effect of curve smoothing on transfer function measurement display, 8k-point FFT; (a) measurement of a
small loudspeaker with no smoothing, (b) the same data displayed with 9-point smoothing.

üQuick Reference
Linear frequency scale FFT data is distributed on a frequency axis scaled on equal intervals per Hz.
Logarithmic frequency FFT data is distributed on a frequency axis scaled on equal intervals per octave,
scale which corresponds to human pitch perception.
Fractional-octave Used to represent Spectrum mode signals in logarithmically-distributed bands, much
banding like a hardware RTA. Bands may be 1-, 1/3-, 1/6-, 1/12-, or 1/24-octave wide.
Fixed-Point Per Octave Performs Transfer Function mode analysis on logarithmically-spaced (1/24-octave)
(FPPO) frequency bins. Enhances trace readability and correlation to hearing perception
when equalizing systems.
Trace smoothing Further smooths Transfer Function mode signal traces for improved curve readability.
FFT Frame Averaging to Improve Trace Validity

While the signal for each FFT frame is viewed only for a finite amount of time (the length of the time
constant, TC), we can modify the time response of the measurements by using frame averaging
techniques. Effectively, averaging causes the current displayed FFT data to reflect some cumulative
average between the current data and past FFT frames. By varying the depth and function of the
averaging, we can achieve changes in the apparent time response for measurements.
In SmaartLive’s Spectrum mode, several options are available for configuring the averaging technique
used for measurement. The spectrum mode contains options for FIFO (First-In/First-Out) averaging,
where the last 2, 4, 8, 16, 32, 64, or 128 FFT frames are averaged with equal weighting, and the result is
displayed (see Figure 6). These modes allow the user to fine-tune the averaging depth depending on the
measurement task and input signals: higher averaging frame counts provide a slower time response, with
lower numbers better approximating the instantaneous behavior of the signal. The spectrum mode also
contains exponential averaging techniques, which are marked Fast, Slow, and Exp. The Fast and Slow
settings reflect the standard damping of a sound level meter in the associated integration mode, which are
most useful for performing repeatable, standardized measurements. The Exp mode allows for the
exponential averaging half-life to be customized, allowing the exponential averaging modes to be user-
optimized based on the measurement task. Finally, the Inf setting allows for an (effectively) infinite-
length average of the input data, which provides a running, equal-weight average of all FFT frames since
the last buffer reseed. This is useful for general noise-level analysis tasks and specialized uses in cinema
optimization, etc. It should be noted that the spectrum-mode averaging is performed on the power
spectrum of the input signal.
4-Frame FIFO Buffer

1 2 3 4
Latest frame Oldest frame
Displayed
average
Figure 6: 4-frame FIFO-based averaging example; the displayed curve is a function of the mean curve from the last N
FFT frames.

Similarly, SmaartLive’s Transfer Function mode contains averaging capability, which provides two
primary functions: (1) it provides a variable time response (identical to the behavior seen in spectrum
mode), and (2) it provides a mechanism for increasing the validity and signal-to-noise ratio of the transfer
function measurement. By averaging transfer function measurement frames over time, the estimate of the
system transfer function increases in precision, with any random noise effects being reduced. When using
the FIFO-based averaging scheme, a doubling of the number of averages corresponds to a 3 dB increase
in the signal-to-noise ratio of the measurement. In addition to basic averaging depth options, the transfer
function mode allows the user to select either RMS or Vector averaging types. The difference between
these two modes is the transfer function measure that is being averaged; in RMS mode, the geometric
mean of the transfer function magnitude is computed, whereas in Vector mode, the complex transfer
function (including phase) is used for the averaging operation. Vector mode provides the most robust
averaging technique for use in phase/time-invariant situations, where no wind gradients or rapid
variations in the measured field are present. For most common measurements, however, RMS averaging
provides a reasonably effective method of increasing the validity of the transfer function measurement
with less sensitivity to these issues. It should be noted that, in RMS mode, the averaged phase trace is
derived from a vector average; only the magnitude curve is obtained through RMS averaging.
SmaartLive’s Impulse mode utilizes an RMS-based averaging technique, which operates solely on a
FIFO-style FFT frame buffer. As with the transfer function mode, increasing the averages by a factor of 2
corresponds to a 3 dB increase in signal-to-noise ratio. The sole purpose of averaging in the impulse
response mode is to improve data validity and reduce the impact of measurement noise, although higher
averaging depths will directly correspond to longer acquisition times in this mode.
üQuick Reference
Frame averaging Improves data validity and trace stability by deriving the displayed data from a
running average of the current data and past FFT frames.
FIFO averaging The displayed data is an equal-weight average of the current input data with a finite
number of past frames. Available FIFO lengths are 2, 4, 8, 16, 32, 64, or 128 frames.
Exponential averaging Provides an integrated averaging response similar to that of a sound level meter.
Provided are Fast, Slow, and Exp (custom) response characteristics.
RMS averaging Averaging behavior in this mode utilizes the geometric mean of the response data.
Used in the Spectrum and Impulse modes and available as a option in the Transfer
Function mode. Useful in Transfer Function mode when the system is varying in time
(wind gradients, etc.).
Vector averaging Averages the complex value of the transfer function data, providing maximum
precision when the system is time-invariant.
Measurement Time Windows

We have noted previously that, for real-world measurements, the FFT is only able to observe the signal
for a finite amount of time, which we may call the measurement time window. Beyond the basic
observation that we may only draw conclusions about the signal from this narrow slice of time, this
operation also creates anomalies in the frequency domain, which may lead to errors in the observed
frequency spectrum.

Time Domain Frequency Domain
amplitude dB
-8 +8
(a)
time frequency
amplitude dB
(b)
time frequency
amplitude dB
(c)
time frequency
Figure 7: Time windowing effects in the Fourier transform with a sinusoidal signal; (a) an infinite time window, (b) a
finite Rectangular time window, (c) a finite Hanning time window.
Figure 7 demonstrates the effect of a finite time window on the FFT spectrum of a sinusoidal (single-
frequency) input signal. A steady-state sinusoidal signal has energy at only one unique frequency, so the
ideal spectrum produced by the Fourier transform should indicate this by an infinitely narrow “line” at
that frequency. As shown in Figure 7a, the Fourier transform is able to accurately determine the
spectrum, given an infinite time record of the signal. However, if we reduce our view of the signal to a
finite-length FFT frame, we, in effect, multiply the time signal by a rectangular window, as shown in
Figure 7b. The effect of this windowing operation on the frequency domain is obvious, as the energy is
dispersed into a single main lobe plus a number of side lobes (called spectral leakage). The pattern of the
side lobes is a function of the rectangular window, and is primarily created by the abrupt discontinuity at
the edges of the window. The high relative level of these side lobes could cause other important spectral
information to be hidden, or masked. However, if we apply a finite but gentle time windowing function,
as seen in Figure 7c, we can significantly reduce the level of the side lobes by gradually tapering the
waveform to zero at the ends. A side effect of this operation causes the main lobe to increase in width,
which may also mask spectral details from nearby frequency components.
We can balance the side lobe magnitude and pattern against the main lobe width by careful selection of
the windowing function, which can be customized depending on the analysis task. SmaartLive contains a
number of time windowing functions which provide varying characteristics in time and frequency. For
more information, see the SIA technical note “The ‘Mystery’ of Data Windows”, which covers this topic
in further detail.
Optimal FFT Parameter Settings for SmaartLive

This section will provide suggestions for reaching the optimal FFT parameter settings depending on the
measurement task at hand. As each parameter is highly configurable, a general set of criteria is presented
here that allow you to optimize the measurement system depending on the desired temporal and spectral

resolution. For most uses, once an optimal set of parameters is reached for a specific measurement task,
the FFT parameters may be left unmodified for subsequent measurements.
A parameter that will not be specifically addressed as a variable is the selection of an optimal sampling
rate, which defines the highest measurable frequency (Nyquist) and the time resolution of the
measurement. As noted previously, you should typically select the highest compatible sampling rate
between SmaartLive and your hardware interface, usually either 48 kHz or 44.1 kHz. On rare occasion
this guideline may be modified, such as those situations requiring a lower-bandwidth measurement where
computational power may be conserved by choosing a lower sampling rate.
Spectrum Mode
For most real-time analysis measurements in Spectrum mode, we are

primarily concerned with three major criteria: appropriate spectral resolution
for the target measurement, an appropriate response characteristic in time, and ü Spectrum Mode
You may wish to begin with

a reasonable correlation to human hearing perception. We may use these these parameter settings:
basic criteria to develop guidelines for choosing FFT parameters in this mode.
In spectrum mode, the FFT size should be configured for an optimal trade-off
between low-frequency resolution and time response. A good starting point
may be to choose a 16k-point FFT frame, which tends to provide acceptable
frequency resolution with an ~350 ms time constant (at SR = 48 kHz) for a
reasonable time response characteristic. For detecting dynamic changes in
signals, a Fast averaging characteristic is useful, with the Slow setting more
useful for analytical measurements (noise, etc.). Of course, you may
experiment with different FFT parameter settings to optimize performance.
For most measurements in this mode, you will wish to choose a banded
display; a 1/3-octave display is a good starting point for many measurements,
as it corresponds to the critical bandwidth of hearing perception for most
complex signals. Higher resolution displays may be used to more accurately detect spectrum details, such
as feedback center frequencies, etc.
Transfer Function Mode ü Transfer Function Mode
You may wish to begin with these

parameter settings:
Most transfer function measurements fall into one of two categories:
measurement of the electrical transfer function of a device (an
equalizer, etc.), or measurement of the electroacoustical transfer
function of a complete system (a loudspeaker, sound system, etc.).
For most electrical measurements, choosing a standard linearly-
spaced FFT is typically a good starting point. Note that, for proper
measurement of low-frequency filters, the FFT parameters chosen
should have sufficiently narrow frequency resolution (start with 16k-
or 32k-point FFTs). Electrical measurements typically provide an
inherently high signal-to-noise ratio, so a short averaging time (8-16
averages or the fast response characteristic) may be used for a more
dynamic time response.
Electrical Acoustical
measurements measurements

When performing electroacoustical measurements, the FPPO measurement mode should typically be used
for best results. FPPO will best approximate the hearing perception of spectral effects in most
electroacoustical systems and is typically far easier to interpret than standard FFTs due to the natural
fractional-octave distribution of data. Electroacoustical measurements are far more susceptible to noise
and time variance effects, so the use of a long averaging time is recommended (64+ averages or the slow
response characteristic). In addition to these parameters, the transfer function mode contains coherence
and magnitude threshold settings, which may used to ensure valid data. Typical settings for these
parameters when performing electroacoustical measurements are: coherence threshold = 10-15%, and
magnitude threshold = 35-55%. The details of these parameters are outlined in following articles, and in
the SmaartLive User Guide.
Impulse Response Mode
Selecting appropriate FFT parameters for the Impulse mode is somewhat

more straightforward than for other modes in SmaartLive. Two primary
criteria are at work for impulse response measurements: the decay time of the
ü Impulse Mode
You may wish to begin with

measured impulse response and the effective signal-to-noise ratio of the these parameter settings:
measurement. For effective impulse response measurements, the time
constant TC must be greater than the total decay time of the system under test
in order to assure valid data.
In performing impulse response measurements, it is often difficult to obtain

an acceptable signal-to-noise ratio in order to fully view the entire decay of
the system. As with transfer function measurements, increasing the number
of averages will increase the effective signal-to-noise ratio of the
measurement. An identical effect is seen by simply increasing the time
constant; this provides an equivalent increase in the time the system is
Note: Increase FFT size such
observed in order to increase the signal-to-noise ratio. Other methods exist that TC exceeds the system
for lowering the noise floor, such as increasing the acoustic level of the decay time. You may increase
stimulus or by carefully selecting the stimulus signal type (such as using the FFT size, averaging depth,
and the stimulus level to
synchronized sinusoidal sweep signals). These advanced concepts are maximize the signal-to-noise
discussed in subsequent articles, as well as in the SmaartLive User Guide. ratio of the measurement.
Suggestions for Further Reading

If you have an interest in furthering your knowledge of the engineering concepts behind FFT-based
measurement, you may find the following texts useful:
A. Oppenheim, A. Willsky, S. Nawab: Signals and Systems, 2nd edition. Upper Saddle River, NJ:
Prentice Hall Inc. 1997.
A. Oppenheim, R. Schafer, J. Buck: Discrete-Time Signal Processing, 2nd edition. Upper Saddle
River, NJ: Prentice Hall Inc. 1999.

NOTES
Spectrum Measurements with SmaartLive:

Concepts and Applications
Presented by:
All rights reserved
Spectrum Measurements with SmaartLive:
Concepts and Applications
Paul D. Henderson
A key component of the SmaartLive application is its Spectrum mode, which provides a highly flexible
FFT-based real-time analyzer capable of advanced audio signal measurements. The spectrum mode is
useful in many applications, including noise and sound exposure measurements, the location of feedback
frequencies in sound reinforcement, and cinema system optimization, as well as general signal monitoring
tasks. This document serves as an introduction to effectively applying the Spectrum mode measurement
capabilities. A general knowledge of the topics covered in “Getting Started with SmaartLive” and “The
Fundamentals of FFT-Based Audio Measurements in SmaartLive” will be helpful when reading this text.
I. Basic Spectrum Analysis

The real-time analyzer, or RTA, is a familiar tool to audio professionals, allowing the user to view the
spectral content of audio signals throughout a system. Traditionally, RTAs were constructed as dedicated
hardware devices, capable of displaying signal power versus frequency in fractional-octave bands. In
early devices, banks of analog filters separated the spectrum into constant-percentage bandwidth sub-
bands, and a level detector circuit was used to drive the unit’s display. The result was a limited-resolution
frequency domain view of the spectrum of the incoming signal (Figure 1), akin to the limited-time view
provided by an oscilloscope.
Amplitude
Frequency
Figure 1: Traditional real-time analysis: fixed & limited resolution.
SmaartLive provides an advanced alternative to traditional RTAs, taking advantage of the computational
efficiency afforded by the Fast Fourier Transform (FFT) and digital signal processing. SmaartLive has
the capability to provide a much higher spectral resolution than filter-based approaches, and can
simultaneously acquire a temporal signature of spectral changes over time. SmaartLive can display the
fine structure of the linear-frequency FFT spectrum of the input signal or may be configured to show a
fractional-octave (banded) view with similar response to a hardware RTA.
This section will introduce the various measurement techniques used throughout the Spectrum mode,
specifically targeting those measurement tasks that do not require an absolute-level calibration. Several
applications are reviewed, including monitoring and troubleshooting of audio equipment and the location
of electroacoustical feedback frequencies.
Spectrum-Mode Measurements with SmaartLive: Concepts and Applications Page 1

Monitoring Audio Signals
For this application, we will investigate the use of the Spectrum mode to view the frequency content of
generic audio signals, such as those received by a microphone or output by a mixing console. Figure 2
shows a simple measurement configuration for this task, where SmaartLive monitors the output of both a
microphone preamplifier and a mixing console. While a simple configuration, the connection diagram in
Figure 2 is a powerful tool for use in sound reinforcement, as SmaartLive may be instantly switched
between the Spectrum and Transfer Function modes for quick access to information for both equalization
purposes (the transfer function) and the live signal spectrum.
Measurement
Microphone
Microphone Line
Preamplifier Input
L
Computer with
R SmaartLiveTM
Mixing Console
Figure 2: Monitoring microphone and mixing console signals.
Basic RTA Use
If we launch SmaartLive with its default parameters

and start the Spectrum mode, we will see a graph
representing the signal power-versus-frequency in
1/12th-octave frequency bands for both input
channels (see Figure 3). You will note that the time
response of the RTA display is quite fast; this is due
to the short time constant (186 ms) and the fact that
there is no frame averaging applied to the display
(averages = 1). A fast time response is useful only
for very specific applications, primarily when a joint
time-frequency analysis of a signal is desired, such
as with a spectrograph (as we will see later). This
short time constant also provides minimal low-
Figure 3: Default live RTA spectrum display (1/12th-octave).
frequency resolution, where the frequency resolution Green curve: microphone, blue curve: mixer output.
in this case is
ü Technical Note
The exponential averaging modes in

5.4 Hz. As you can see in Figure 3, the low-frequency bands seem
to have a smooth appearance, due to the low density of FFT data
bins at these low frequencies.
SmaartLive conform to the ANSI S1.4-1983
standard, where the exponential time
constants t are as follows:
To improve the spectral view, we can increase the FFT size and use
frame averaging to adjust the time response of the display.
Fast: t = 125 ms Changing the FFT size to 16k-bins and the averaging to slow
Slow: t = 1000 ms

provides both additional low-frequency resolution and a more stabilized time response.
The slow characteristic is based on a running exponential integration, and is useful for
continuous signal monitoring during a live performance, noise level measurements, etc.
The fast characteristic provides a much more immediate time response, and is useful for
viewing transient signal information, such as specific musical notes. Note that the
spectrum mode also allows for FIFO-style frame averaging, but, in most cases, this
averaging technique is seldom required or recommended for Spectrum mode
measurements. Note that both banded (1/24th, 1/12th, 1/6th, 1/3rd, and 1/1-octave) and
narrowband (linear- and log-spaced) views are available. For a more detailed treatment of
the FFT, averaging, and banding parameters, please see the article “The Fundamentals of
FFT-Based Audio Measurements in SmaartLive”.
You will note that, by default, the Spectrum display shows two
separate traces, one for each input channel. SmaartLive contains a set
of simple controls for managing the display of the two input channels,
located directly underneath the input signal level meters. You may
define either of the two input channels as the active channel by
clicking the active indicators below the level meter. The current active
trace will be brought to the front of the spectrum display, and the SPL
Meter, SPL History Graph and Spectrograph displays will reflect the
input signal from this channel only. The color of each of these Set active
displays will change to reflect the currently active channel. When trace
using these advanced functions, it is important to note that the
appropriate input channel is selected as active in order to obtain valid
data. In addition, clicking the show/hide channel buttons will enable or
Show/hide
disable the display of the associated spectrum in the RTA window. traces
By default, traces in the RTA view contain peak hold bars, which hold the maximum value of the
measurement parameter at each band or bin for a specified period of time. The configuration parameters
for the peak hold mode may be found in the Input Options dialog box, which may be
launched with the menu command Options? Input, or by pressing the key combination
Alt+I. The hold time can be customized or an infinite hold mode may be activated,
which holds the detected peak levels for an indefinite period of time until the next buffer
reseed (performed by pressing the V key on the keyboard). In addition, the peak hold
bars may be removed from the display for a more continuous, orderly appearance.
By hovering the mouse pointer over the RTA spectrum, the level and
center frequency of the selected band is shown in the readout above
the trace display. In addition to this function, SmaartLive can display the musical note
associated with the band center frequency using Note ID mode. Note ID mode may be toggled
using the View? Note ID menu item. In addition, a reference piano keyboard may be
activated using the Graph Options dialog box, accessible under menu item Options? Graph.
You may use SmaartLive’s reference trace functions to save a snapshot of the measured
spectrum to a storage register for later recall and on-screen comparison. SmaartLive contains five
reference registers (A, B, C, D & E), each capable of storing up to four individual traces. As an example,
clicking the button for register A1 (the first register button in the A group) will activate
this memory location. Click the button below the plot area to sample and display the
current trace as an overlay on the plot. Note that the curve that will be captured is the
current active trace, as defined above. The reference register stores the raw, post-

averaging FFT data, so the curve may be later viewed using any desired fractional-octave banding or
narrowband display scheme.
Viewing a Long-Term Spectrum Average
The averaging buffer may be set to Inf, which will

provide an equal-time-weighted average of the entire
signal spectrum since the last buffer reseed. This
method can provide you with a view of the long-
term spectrum of the signal, a measure of the
“spectral signature” of the input signal. For
example, the 1/3rd-octave spectrum in Figure 4 is a
long-term average of a musical performance. This
technique is also useful for capturing a stable
measure of the continuous noise level in a room,
which will be discussed later. Note that pressing the
V key on the keyboard will reseed the averaging
buffer, restarting the infinite averaging procedure at
Figure 4: 1/3rd octave long-term spectrum of a musical
any desired time. performance.
Performing Harmonic Distortion Estimates
An interesting and sometimes-overlooked feature of SmaartLive’s Spectrum mode is its capability to

perform a rough Total Harmonic Distortion or THD measurement at a single frequency. THD is a metric
used to quantify the nonlinear distortion of a system or device. A system with nonlinear distortion will
output multiple frequencies for a single-frequency (sinusoidal) input, where those output frequencies will
be multiples (or harmonics) of the input signal. This is typically due to a nonlinear “transfer
characteristic” between the system input and output, which can be caused by tape saturation, digital
clipping, transistor saturation, or nonlinear effects in loudspeakers. SmaartLive’s THD capability is
useful for quantifying the distortion measure of analog recording devices and power amplifiers, as well as
for determining a (rough) distortion metric for loudspeakers.
For this measurement, you will need to use

SmaartLive’s internal generator to output a
single-frequency tone into the device-under-test
at a reference frequency, and a low-distortion
audio interface. The measured THD will
represent the total distortion of the entire
measurement system, so it is imperative that the
measurement system have significantly lower
distortion than the device under test. Figure 5
shows an example configuration, where a THD
measurement of a loudspeaker is being
performed. By placing the RTA in 1/24th-octave
resolution mode and activating the menu item
under View? Cursor? Show THD,
SmaartLive will calculate the total harmonic Figure 5: Sample THD measurement of a small loudspeaker at
distortion from the amplitude of the fundamental 200Hz: THD=0.12%.
and subsequent harmonics at the current cursor

position. Set the generator to Sine Wave to output a single frequency of choice, and place the cursor at
that frequency. It may be necessary to increase the FFT size to maximize the low-frequency resolution of
the measurement. Figure 5 shows a sample measurement of a small loudspeaker at 200Hz. For the
measurement of a loudspeaker, the room should be very quiet, and a low-distortion microphone should be
employed. Note that you may also use the Show Harmonics feature of SmaartLive by pressing Ctrl+H,
which will activate a grid marking the harmonics relative to the locked cursor position. See the
SmaartLive User Guide for more information.
II. Viewing Spectral Changes in Time: the Spectrograph

Many audio signals that are encountered in the field are highly dynamic: musical signals, speech, and
even environmental noise contain significant changes in spectral content as a function of time. While the
classical RTA spectrum display is useful for viewing the time-averaged spectral content of a signal, a
display technique is required to enable the analysis of the incoming spectrum over time. SmaartLive’s
Spectrograph mode fulfills this requirement, providing a three-dimensional view of the incoming signal:
energy versus frequency versus time (Figure 6). The spectrograph can be thought of as a record of
multiple RTA spectrums taken over time, with color representing amplitude. Using this functionality, the
spectral content of the input signal is recorded as it changes in time, allowing the user to view and analyze
time-varying trends in the input signal. The Spectrograph display is launched while in Spectrum mode by
pressing the button. Note that the Spectrograph display is derived from the current active trace.
Figure 6: An example spectrograph of a musical performance.
As a troubleshooting tool, the Spectrograph mode is useful for finding spectral “defects” in a system or
acoustical environment. Certain audio signals or acoustical events contain specific traits that can be
easily detected due to their distinct time/frequency signature, specifically, highly tonal sounds such as AC
line noise in an electrical signal chain or the presence of electroacoustical feedback.
Configuring the Spectrograph
In order to achieve optimal results with this measurement mode,

the Spectrograph configuration options must be properly set up
for the measurement task at hand. The input banding, color
range, and time averaging options are critical to achieving
readable displays with this mode. The Spectrograph Options
dialog may be launched using menu item
Options? Spectrograph or by clicking the Spectro Range scale
indicator. This dialog allows you to configure the color scale
and amplitude range as well as the frequency and time axis

behavior. When configuring the spectrograph, the following issues are important to note:
• Spectrum banding: The banding scale of the spectrograph display tracks the selected banding of
the RTA. For most applications, a 1/24th-octave or log-narrowband scale provides the best
broadband frequency resolution. Low fractional-octave resolutions typically result in a rough
appearance.
• Time response: By default, the time response of the spectrograph display tracks the averaging
depth selected for the RTA. However, excessive averaging (or averaging at all) can obscure
time-based details in the spectrograph, such as in Figure 7b. For most applications, check the
Instantaneous Spectrograph box in the Spectrograph Options dialog. With this option, each
spectrograph frame will represent a minimum-width time slice of the input signal, improving the
time response of the display (Figure 7a). In addition, the number of FFT frames contained within
the spectrograph may be set in the Frames to show in Spectrograph field. Higher numbers yield a
slower time response, with more time history contained within the window. Conversely, lower
numbers produce faster movement along the time axis. The absolute-time length of the time axis
is a function of this value and the speed of your computing hardware.
• Amplitude range: It is imperative that an appropriate amplitude range be selected in order to
reveal the desired amplitude details from the input signal. The current range is configured using
the Min dB and Max dB boxes in the options dialog. Selecting too large an amplitude range may
obscure amplitude details by plotting a minimal change in color as the input spectrum changes
(Figure 7c), and too narrow a range or values too high or low may lead to undershoot and
overshoot errors (Figure 7d). The selection of an appropriate scale will enable the detection of
spectral details with high definition in amplitude, as in Figure 7a.
See the SmaartLive User’s Guide for information regarding additional configuration options for the
spectrograph display.
(a): Instantaneous spectrograph, acceptable magnitude (b): Slow-weighted time response,

range calibration & good time resolution time-domain details lost
(c): Magnitude range too wide, (d): Signal exceeds maximum magnitude range limit,
amplitude details lost peaks lost in clipping (white regions)
Figure 7: The effect of selecting various spectrograph parameters;

1/24-octave banded spectrograph of an identical input signal.

Locating Feedback Frequencies in Sound Reinforcement
Here, we will discuss the use of SmaartLive in determining electroacoustic feedback frequencies in sound
reinforcement and/or stage monitoring systems. The location of feedback frequencies during live sound
performance is a task especially well-suited for the Spectrograph mode, which provides an advantage to
detecting the unique time/frequency signature of feedback ring tones.
Microphone Stage Monitor
Acoustical Environment
Power
Equalizer Preamplifier
Mixing Console
Line
Input
L
Computer with
R SmaartLiveTM
Figure 8: Finding feedback frequencies in a stage monitoring system.
Figure 8 shows a simplified stage monitoring system with SmaartLive configured to monitor the console
output signal before the equalizer. On a practical note, this connection can be made on the console’s
wedge output with access to the solo bus, so that any monitor output send can be retrieved immediately.
A typical optimization process for a stage monitor system would entail an accurate equalization process
using SmaartLive’s Transfer Function mode, followed by in-show dynamic equalizer changes based on
feedback modes that may be found as the acoustical environment varies. Note that a feedback mode in an
electroacoustical system needs two components to begin “ringing” (or regenerating): a system-borne
instability at any frequency and ambient sound energy at that frequency to initiate the regeneration.
Therefore, it is imperative to note that using any method to locate feedback frequencies (commonly
known as “ringing out” a system) without active signals in the system (such as a live performance) will
result in inaccurate adjustments. If this process is performed without ambient sound such as in a pre-
show configuration, some broadband energy should be introduced to the system (such as low-level pink
noise). Electroacoustical feedback is created by narrowband loop-path instability, and is primarily a
single-frequency effect, although multiple feedback modes may be excited simultaneously.

Feedback Signature
Figure 9: Feedback signatures in spectrograph and spectrum displays.
This time-domain element of the spectrograph can be exploited in the task of finding feedback
frequencies, since the ringing of electroacoustical feedback has a time-domain signature, i.e., it is constant
in frequency and has both a duration and envelope in time. Narrowband constant-frequency signals such
as a feedback ring tone or tonal noise source show up as horizontal lines in the spectrograph, making it
somewhat easier to find these effects in the presence of a more complex signal, such as a live musical
performance. Figure 9 illustrates this effect, as well as the simultaneous use of the spectrograph and RTA
displays. An additional advantage of this technique is that the operator can immediately reduce the gain
to eliminate the feedback loop on the console, and the running spectrograph will contain a history of the
ring frequency for subsequent equalization changes. Note that, when controlling compatible external
signal processors, their control parameters may be changed without leaving Spectrum mode. A Control
frequency response display will appear, allowing you to view calculated EQ traces for compatible devices
just as in the Transfer Function mode.
III. Performing Calibrated Measurements

For the measurement procedures introduced in Section I, SmaartLive was in its uncalibrated state: the
values displayed were relative to the maximum full-scale range of the analog-to-digital converter.
However, some measurement procedures require absolute calibration, where the displayed values are
calibrated relative to real-world physical quantities. While SmaartLive can be calibrated to any decibel-
based scale (dBu, dBm, etc.), the most commonly-used absolute scale is dB-SPL.

Calibrating the System for SPL Measurement
Measurement Microphone Line

Calibrator
Microphone Preamplifier Input
L
Computer with
R SmaartLiveTM
Figure 10: The SPL calibration process.
In Spectrum mode, the entire system can be calibrated in a single step,

including the built-in sound level meter. Several techniques are outlined in
the User Guide for performing the calibration, but we will cover only the
ü
Technical Note
0 dB-SPL = 20 µPa RMS

most accurate and widely-accepted procedure, which utilizes a sound level 94 dB-SPL = 1 Pa RMS
calibrator as an absolute level reference (see Figure 10). The calibrator 104 dB-SPL = 3.2 Pa RMS
should be correctly sized for your measurement microphone in order to 114 dB-SPL = 10 Pa RMS
ensure an airtight seal with the capsule. Most calibrators output a 1 kHz
reference tone at one of the following reference levels: 94, 104, or 114 dB-SPL. Choose an appropriate
reference level and adjust your microphone preamplifier gain to achieve sufficient headroom with this
reference pressure. Place SmaartLive into 1/3rd-octave RTA mode. The 1/3rd
octave mode is employed to make certain that all of the energy from the windowed
sinusoid is taken into account. If calibration is performed in a narrowband or high-
resolution mode, the time window effects may distribute a significant portion of the
sinusoid energy into surrounding bands. Select the input channel from the
microphone as the active channel. Next, double-click on the 1 kHz band (or other
band if your calibrator uses a different center frequency), and the Amplitude
Calibration dialog will open. Select “Set this value to...” and enter the level of your
calibrator. The system is now calibrated to Sound Pressure Level for the current
channel only, unless a microphone and preamplifier with identical sensitivity and gain are connected to
the opposite channel. Note that any change in the preamplifier or sound card gain will necessitate
recalibration. If the clip indicators in SmaartLive illuminate during use, reduce the preamplifier gain and
recalibrate for additional headroom.
Measuring Sound Pressure Level
Now that the system is calibrated to SPL, we can perform

accurate sound level measurements in any environment.
You will note that the digital readout above the input level
meters now displays dB-SPL with a Fast integration time.
By clicking on the readout, you will be taken to the SPL
Readout Options dialog, which enables you to configure
the SPL meter component with various weighting curves
and integration speeds, as well as configuration options
for peak hold, SPL alarms, and logging capabilities. The
peak hold and alarm functions are quite useful for live
sound and noise level monitoring tasks, helping to maintain compliance
with noise ordinances and safety regulations. The following sections
will discuss the available weighting curves and integration speeds and
their appropriate uses.

Weighting Curves
For measuring sound pressure levels, SmaartLive includes three available weighting curves. A weighting
curve is a filter response that is placed in-line with the SPL meter detector before the displayed value is
calculated; the filter allows the response of the detector to vary as a function of frequency. SmaartLive
contains Flat, A and C weighting filters, whose frequency response curves are shown below in Figure 11.
The Flat weighting curve is exactly as the name implies: it has a perfectly linear response over frequency,
meaning that the SPL meter will respond equally to equal acoustic pressure at any frequency inside its
range. The Flat characteristic is useful for determining the total sound pressure at a point in space, as
well as for sound power estimation tasks (beyond the scope of this document).
The A and C weighting curves are drawn from the

ANSI S1.4-1983 standard. The C weighting curve
10
contains a slight high-frequency and low-
frequency rolloff, meaning that the SPL meter 0
detector is less sensitive to energy in those regions
of the spectrum. The A weighting curve contains a -10
much more dramatic low-frequency rolloff,

Magnitude (dB)
-20
indicating a primary sensitivity in the region above
1 kHz. Often, the background work that leads us -30
to these specific curves is overlooked: the A and C
curves are approximate inverses of the equal- -40
loudness contours of human hearing at the 40
phon and 100 phon contour levels, respectively. -50 Flat Weighting
A Weighting
The science behind the equal-loudness contours C Weighting
-60
will not be presented here, but, suffice it to say 10
2
10
3
10
4
that the A weighting curve approximates the Frequency (Hz)
response of human hearing at low levels, and the C Figure 11: SPL weighting curves.
weighting curve does the same at high levels.
Most noise level measurement standards have taken the A weighting function as customary, as, even at
high levels, it more correctly approximates both the annoyance level of the noise and the potential for
hearing damage. It should be stated that the suffix dBA is commonly attached to sound pressure readings
in the A weighting scale, and similarly dBC for the C weighting scale. Flat-weighted measurements are
typically suffixed with dB-SPL.
Integration Speeds
SmaartLive contains three integration speeds for the SPL display: Inst, Slow, and Fast. The Inst setting
effectively corresponds to the Impulse setting on a typical high-end sound level meter, as it displays the
latest SPL data from the sound card with no averaging integration. This setting can be combined with the
Peak Hold function to capture the peak sound level in an environment, specifically where there are
impulsive and transient noise sources. The Fast and Slow settings are ANSI-standard exponential
integration curves. The Fast setting is useful for finding standardized peak sound levels, again, best with
the Peak Hold function or in conjunction with the SPL logging plot (to be discussed later). The Slow
setting is perhaps the most useful for typical measurements, as it provides a longer integration time for
more accurately representing the average sound level in the environment. The Slow response is also
widely accepted in standards for studies in sound exposure and noise pollution (such as during a live
performance).

SPL Time History and Logging Sound Exposure Data
While SmaartLive is capable of monitoring SPL in real time, many measurement tasks require a more
thorough investigation using the time history of sound levels in the environment, specifically when there
are dynamic changes in the sound pressure. Many sound fields that humans encounter are highly
dynamic, such as musical performance, industrial noise, etc. SmaartLive contains several tools to assist
the user in this area, which include a run-time SPL History graph and logging functions capable of
documenting overall sound pressure levels, spectrum histories, and SPL statistical metrics (LEQ, LX, etc.).
This section will briefly overview a few applications for these capabilities.
Using SPL History
SmaartLive’s SPL History function (Figure 12) is useful for real-time monitoring tasks, particularly
during live musical performance. It is simply a running history of the SPL meter display on a per-frame
basis (not including peak hold), and is directly affected by the weighting and integration speeds selected
for the SPL meter. The SPL History display also shows the alarm threshold levels that have been preset
for the SPL meter. You may configure the effective speed of the display by setting the number of shown
frames in the options dialog by accessing the menu item Options? SPL History.
This display is highly useful for monitoring sound levels in live performance, specifically with regards to
maintaining both safe listening levels and compliance to noise ordinances and venue regulations. The
detection of peaks is also very valuable, since the display will hold the peak envelope for a period of time,
making it unnecessary to continuously watch the SPL meter readout.
Figure 12: SPL history display.
Logging Sound Exposure Statistics
In addition to the SPL History functions, SmaartLive contains more advanced sound level statistics
functions enabling the user to determine industry-standard sound exposure statistics. SmaartLive allows
the user to both log SPL data to a file and to perform automated calculation of sound level statistics (LEQ,
LX, LMIN, and LMAX). Here, we will cover the use of the automated statistical functions; documentation
for the simple logging capabilities is provided in the User Guide.
Often in environmental noise studies or sound exposure analysis for live performance, a single-number
SPL value does not sufficiently describe the variations in sound level. However, a complete time history
of pressure levels is surplus information, so a compact, statistical description of the characteristics of the
sound field level is required. Equivalent Sound Level, or LEQ, is the sound level that, if the sound field
were continuous and single-level, would contain the same energy dose over the sampling time as the
measured dynamic field. Percentile Noise Levels or LX, indicate the level that the sound field exceeds for

X-percent of the time; for example, an L90 value of 82 dBA indicates
that the sound level is above 82 dBA for 90% of the measurement
period. SmaartLive automatically calculates the three most common
percentile statistics, L10, L50, and L90. Each of these values is based
on the total duration of observation, which is configurable in the
Sample Period field under the LEQ Setup dialog box. This dialog is
launched through by clicking on the SPL meter display and then
clicking on the Timed Average / LEQ Setup button.
The automated LEQ Log option creates text files of the format shown
in Figure 13. In addition, LEQ analyses may be derived from
Spectrum Log files using the Create LEQ report from log file option,
which also allows for custom percentile categories.
;LEQ
;A Weight
;Date_____ Time____ LAEQ LAMin LAMax LA10 LA50 LA90
06/15/2004 21:36:27 66.3 62.0 81.2 67.9 63.6 62.5
;
;Cumulative Values
06/15/2004 21:36:29 66.2 62.0 81.2 67.8 63.6 62.5
Figure 13: Sample LEQ analysis output file.
Various national standards exist for sound exposure, specifically those for occupational safety, including
those by OSHA and NIOSH. The OSHA-standard thresholds are typically considered somewhat high-
risk, but are reprinted here in Table 1. NIOSH specifically recommends that the Time-Weighted Average
for occupational noise (in effect, LEQ with an observation period of 8 hours) be less than 85 dBA.
Duration per day Sound level

Hours dBA, slow response
8 90
6 92
4 95
3 97
2 100
1½ 102
1 105
½ 110
¼ 115
Table 1: OSHA standard (1910.95) for occupational noise exposure.
Measuring a Calibrated Sound Spectrum
The calibration process not only scales the SPL

meter component to read absolute SPL values, but
also causes the RTA Spectrum display to be
calibrated to absolute SPL. This is particularly
useful for fractional-octave band measurements,
which indicate the total sound power in each band
for advanced analysis tasks (see Figure 14).
Note that, for narrowband signals such as a sine

wave, the chosen fractional-octave bands must be
wide enough to contain all significant energy from
Figure 14: Octave-band spectrum calibrated to absolute SPL.

the sinusoid, including the leakage energy from the time window operation. This is the same reason that
a 1/3rd-octave band scale is recommended for SPL calibration with a pure tone calibrator. You can see
this effect by measuring a sine wave in narrowband Log mode; the peak value of the narrowband plot
will be less than the total power of the input sinusoid. For broadband signals, such as pink noise, you will
note that the individual level of each fractional-octave band decreases as the width of each band becomes
smaller. This is due to the fact that less total energy is contained in each band as the bands become
narrower.
Room Noise Measurement
The calibrated spectrum is useful for

determining room noise levels in consulting
practice and for documenting the background
noise characteristics of a room. Spectrum
measurement techniques are useful for
diagnosing noise level issues in rooms due to
HVAC, exterior noise, motor-induced and
transformer-induced noise, noise from lighting
ballasts and fixtures, etc. Figure 15 shows an
example measurement of the noise radiation
from a set of axial fans. These types of
measurements help determine the cause of
interferences in acoustical environments and
Figure 15: Measurement of axial fan noise radiation; the
provide documentation to enable adequate spectrum peak is the blade-passage frequency.
acoustical corrections to be made.
Noise Criterion Analysis
In addition to basic spectrum measurements,

SmaartLive contains the capability to perform
standard Noise Criterion, or NC measurements
of continuous, broadband noise sources in
rooms. Noise criterion measurement utilize a
family of NC Curves which define thresholds for
specifying the broadband noise level of a room
(see Figure 16). Each curve is designated with a
single number, such as NC-30 or NC-50. NC
rating of the environment is derived from the
maximum octave-band sound pressure level that
meets an NC curve (or an interpolated curve in-
between primary curves).
Figure 16: NC noise measurement; room is rated NC-57.
The shape of the NC curves was originally
developed relative to the level of annoyance of HVAC noise levels in rooms, but has since been accepted
as a standard for quantifying many broadband noise sources. You will note that the NC curves are much
higher in SPL at low frequencies; this is due to the equal-loudness contours of human hearing, given that
the human ear is less sensitive to low frequencies. An important point to note is that NC ratings are not
intended to be used when tonal noise is present; this means that the NC rating of a room is only
significant when the source of the noise is mostly broadband, such as noise from air diffusers, etc.

Cinema System Certification
An application that commonly requires an RTA for measurement is the certification of cinema and home
theater systems to industry standards. While it is recommended that high-resolution equalization and
loudspeaker setup tasks be performed with SmaartLive’s Transfer Function and Impulse Response modes,
most cinema standards groups require an RTA-based (usually 1/3rd-octave) measurement for certification.
This section will not cover the specific standards, but will review specific capabilities in SmaartLive for
performing these measurements.
Timed Spectral Averaging
SmaartLive contains a Timed Spectral Averaging mode that allows

for long-term acquisition of average spectra present in the
environment. This mode is accessed by clicking on the SPL meter
display and then clicking on the Timed Average / LEQ Setup button.
As with LEQ measurements, you may select any observation period
for the measurement. The result of this acquisition process is placed
into a user-defined reference register. Because of the potential for a
very long integration time, Timed Spectral Averaging is valuable for
obtaining highly stable measurement results with random noise
sources, such as the use of pink noise during an equalization process
or for quantifying broadband room noise.
Verifying Filter Calibration
Typical large-format cinema systems utilize the ANSI/SMTPE 202M X-curve for equalization of the
main loudspeaker system. The X-curve, shown in Figure 17, provides a uniform 3 dB/octave high
frequency rolloff beyond 2 kHz, and a (somewhat less important) low-frequency 3 dB/octave rolloff
below 63 Hz. When a cinema system in a large room is equalized to this specification, less high
frequency energy will reach the listeners, which has been found to be more audibly-pleasing than a flat
response when the listeners are seated at a distance from the loudspeaker system.
When certifying these systems for cinema use, the Inverse X-curve may be activated
under the Weight spinner for the RTA. Now, with pink noise excitation, a loudspeaker
system correctly conforming to the X-curve will display a flat 1/3rd-octave spectrum in the RTA display.
5
0
Magnitude (dB)
-5
-10
2 3 4
10 10 10
Frequency (Hz)
Figure 17: ANSI/SMTPE 202M X-curve.

Spatial Averaging
In many cinema standards, spatial averaging of measured spectra is required in order to verify that the
correct performance is obtained across the entire audience area. SmaartLive simplifies this task using
Reference Trace Averaging, which allows several measurements to be performed at different seating
positions and then power-averaged to produce an overall performance curve. In order to average up to
four reference traces, capture each measurement into a separate reference bank (i.e., A1, B1, C1, and D1).
The E bank is used to hold the average; activate the E register, press the button, and the press the
button. The averaged curve will appear on screen with the original reference curves (Figure 18).
Figure 18: Spatial averaging of 1/3rd-octave RTA measurements using the averaging buffers.
Gray curves: individual measurements, Red curve: spatial average.
Suggestions for Further Reading

D. Davis, C. Davis: Sound System Engineering, 2nd edition. Carmel, IN: SAMS. 1994.
M. Mehta, J. Johnson, C. Rocafort: Architectural Acoustics: Principles and Design. Upper Saddle
River, NJ: Prentice Hall. 1999.
American National Standard: Specification for Sound Level Meters, ANSI S1.4-1983. New York,
NY: Acoustical Society of America. 1983.
R. Cabot, B. Hofer, R. Metzler: Standard Handbook of Video and Television Engineering: Chapter
13.3: “Nonlinear Audio Distortion”, 4th edition. McGraw-Hill Professional. 2003.
OSHA Standard: Occupational Noise Exposure, OSHA 1910.95. Washington, DC: Occupational
Safety & Health Administration. 1996.
Criteria for a Recommended Standard Occupational Noise Exposure, Revised Criteria 1996,
DHHS (NIOSH) Publication No. 96. Atlanta, GA: National Institute for Occupational
Safety & Health. 1996.
Motion-Pictures - B-Chain Electroacoustic Response - Dubbing Theaters, Review Rooms and

Indoor Theaters, SMPTE 202M-1998. White Plains, NY: Society of Motion Picture and
Television Engineers. 1998.

NOTES
Loudspeaker Impedance Measurements

with SmaartLive
Presented by:
All rights reserved
LOUDSPEAKER IMPEDANCE WITH SIA SMAARTLIVE®
SIA SmaartLive Technical Note
Paul D. Henderson
Program in Architectural Acoustics, Rensselaer Polytechnic Institute, Troy, NY
A recent addition to the capability of SIA SmaartLive is the ability to measure complex load
impedance as a function of frequency. The potential to perform these measurements permits
investigations of loudspeaker behavior in the field with accuracy previously available only in the
laboratories of loudspeaker manufacturers. This tool may be used to troubleshoot loudspeaker drivers,
systems, and constant voltage networks, as well as to design related systems and select optimal
loading conditions for power amplifiers. This article provides an overview of the measurement
technique and necessary theory for taking advantage of this useful tool.
1. AN INTRODUCTION TO LOUDSPEAKER IMPEDANCE
The term impedance is widely used in the professional audio industry, but frequently misunderstood
and misapplied. Impedance is the total opposition to the flow of alternating current (AC current) in
an electric circuit, and is a complex function of frequency as the ratio of voltage to current (Equation
1). The concept of impedance is analogous to resistance in direct-current (DC) circuits. While
impedance includes resistance, it includes another element exclusive to AC circuits, reactance, which
is due to the energy storage effects in AC circuits from components like inductors and capacitors,
which vary as a function of frequency. In
engineering circles, impedance is thought of as a V( f )
complex quantity, meaning it includes both real Z( f ) = (Eq. 1)
(resistive) and imaginary (reactive) parts (Equation
I ( f )
2). It is this concept that accounts for the varying Z ( f ) = R( f ) + j X( f ) (Eq. 2)
phase shift of impedance: current flows through
resistive components in phase with the applied Z( f ) = Z ( f ) = R 2 ( f ) + X 2 ( f ) (Eq. 3)
voltage, while current flows through reactive
components with a phase shift relative to the applied voltage. The impedance magnitude (Equation 3)
contains the effects of both the resistive and reactive components, and indicates the total opposition
to current in the circuit (ignoring phase). It is this magnitude function that is typically quoted in
loudspeaker specifications, as it is the impedance magnitude that affects the total current required
from an amplifier when driving the loudspeaker. While the above
general concept of impedance is universally used in many circuit
I analysis tasks, the concept of load impedance or the input impedance of
a load (such as a loudspeaker) seen by a driving source (such as a
power amplifier) is what we typically deal with when looking at
V Z loudspeaker characteristics (see Figure 1).
The electrical input impedance function of a real loudspeaker is

defined by many factors, including electrical, mechanical, and
acoustical behavior. Electrically, the resistance and inductance of the
voice coil dominates, along with the presence of any passive crossover
Figure 1: components. The mechanical mass, compliance (“springy-ness”), and
resistance of the drivers form another component of the total
impedance. Additionally, the acoustical impedance seen by the drivers
appears as part of the electrical impedance function, including the baffle loading effects, any
loudspeaker ports, etc. All of these factors combine to create the impedance functions seen by
measuring a typical loudspeaker.
Loudspeaker Impedance with Smaart
Figure 2 shows the impedance magnitude-versus-frequency curve for a single low-frequency driver in
both a sealed and ported enclosure. The strong dependence of impedance on frequency is easily seen.
In the sealed example, the peak is created by the
resonance between mechanical compliance and mass
in the driver. The second peak appearing in the
ported case is the acoustical tuning resonance of the
vent. The rise in impedance at high frequencies is
due to the inductance of the voice coil, while the
minimum value of the graph is equal to the resistance
of the voice coil. These characteristics are typical
examples of measured data from real loudspeaker
systems, providing vital information about the
loudspeaker system for troubleshooting and design.
The included bibliography lists several excellent Figure 2: Input impedance of a single low-frequency driver
in both a sealed and ported enclosure.
references for interpreting and applying this
information.
2. IMPEDANCE MEASUREMENT TECHNIQUES
Using the concepts developed in Section 1, we can now investigate methods of measuring load
impedance. Based on Figure 1, we can see that the load impedance function may be obtained directly
if we are able to acquire signals representing both the voltage and the current into the load impedance
over all frequencies of interest. Since computer sound cards respond to voltage signals, a signal
proportional to the voltage across the load is easily acquired by simply feeding the load voltage
directly to the sound card. However, other techniques must be used to acquire the current signal. The
current signal is most easily measured by inserting a
shunt resistor in series with the load, creating a current
shunt; the current in the load is then directly
I·Rshunt proportional to the voltage across this shunt resistance
(see Figure 3). This is the method employed by most
digital multimeters on the market to measure current.
There are other methods of deriving the current signal,
including the use of inductive current probes, etc.,
Z V however, the shunt resistance method is the most
practical technique for measuring loudspeaker
impedance with SmaartLive.
As outlined above, if the voltage across the shunt

Figure 3: Impedance measurement with a shunt resistor.
resistance is measured, the current may be derived.
However, if the source in Figure 3 is ground-referenced, the shunt resistor voltage is floating, so a
differential amplifier must be used to directly measure this voltage. Using SmaartLive, the circuit
required to use this differential (balanced) measurement technique is shown in Figure 4. The
differential inputs may be provided using anything from a laboratory-grade differential amplifier to a
balanced line-level input on a professional audio interface or mixer. However, for Smaart users
desiring a simpler interface or lacking a differential input, a single-ended (unbalanced) measurement
technique is possible, with SmaartLive calculating the exact load current internally. Both techniques
have advantages and disadvantages, along with trade-offs associated with the selection of the value of
the shunt resistor.
Page 2 of 7
Figure 4: Smaart impedance measurement circuit with a Figure 5: Smaart impedance measurement circuit with a single-
differential (balanced) input technique. ended (unbalanced) input technique.
Table 1 compares the two measurement techniques, presenting the trade-offs associated with each. In
general, for a laboratory-grade measurement solution, choose the differential method with a high-
grade balanced-input preamplifier. If you desire a simple, practical solution, choose the single-ended
method, being certain to adequately calibrate your measurement system appropriately. SmaartLive
requires you to use a calibration resistor to calibrate the measurement configuration based on this
reference resistance for maximum accuracy. This calibration resistance temporarily replaces the load
impedance during the calibration routine, which will be reviewed later.
Differential Method Single-Ended Method

Advantages: Advantages:
• High-accuracy, large-range with low noise • Simple, passive design
differential amplifiers • No common-mode considerations
• Insensitive to differences in input channel
gains; simple calibration
Disadvantages: Disadvantages:
• Complex input circuitry required • Sensitive to mismatch in input channel
gains; requires careful calibration
Table 1: Comparison of impedance measurement methods.
3. OPTIMIZING THE MEASUREMENT HARDWARE
Before you jump in and start making impedance measurements, you should be aware of the practical
issues involved in building an impedance measurement circuit and interfacing this device with your
computer. Selecting appropriate shunt and calibration resistors will affect the quality of your
measurements, and care must be taken when interfacing the computer with loudspeaker-level signals.
Driving the Circuit
Just like transfer function and impulse response measurements in SmaartLive, the impedance
measurement function is dependent on a broadband excitation signal, such as random noise or a
sinusoidal sweep, to perform its measurement. While transfer function measurements will use this
signal to drive input of the device under test, the impedance function uses this signal to excite the
load through the current shunt.
When a dynamic loudspeaker is driven by a source with relatively high source impedance (like a
current shunt), the effect of the loudspeaker acting in reverse as a microphone may affect the quality
of your measurements if there is sufficient acoustic noise in the measurement room. In order to
minimize this problem in situations with a high ambient noise level (such as a construction site,
manufacturing floor, etc.), the shunt resistance must be relatively small, and a small power amplifier
used to drive the circuit. This, however, must be carefully undertaken in order to avoid damage to
Page 3 of 7
your computer hardware. Power amplifier signals can easily exceed the maximum input voltage
capability of conventional computer sound cards, and the power handling capability of the shunt
resistor becomes increasingly important as more current is drawn into the load. When using a power
amplifier, the output voltage must be limited to a safe level and power resistors used in designing the
current shunt.
In situations where ambient noise is not a concern (closed sites, laboratories, etc.), the headphone
amplifier of the sound card may be safely used as the excitation source. Headphone amplifiers are
typically rated for load impedances greater than ~30Ω, so the shunt resistance should be at least equal
to this value. For typical loudspeaker loads, this shunt resistance limitation does not pose any
question of accuracy to the measurement as long as ambient acoustic noise levels are minimal.
Selecting Resistors
The value, precision, and power handling capability of the resistances should be optimized when
configuring the circuit based on the measurement conditions and load. Selecting an appropriate shunt
resistance allows you to optimize the signal-to-noise ratio of the measurement while best taking
advantage of the dynamic range and resolution of the system. Conversely, the calibration resistor
should be of high precision and on the same order of magnitude as the unknown load impedance for
maximum accuracy. For designs using a power amplifier, 2%-tolerance non-inductive wire wound
resistors may be safely used, which are readily available with power ratings ≥10W. In current shunt
designs where a headphone amplifier is used, 1%-tolerance metal-film resistors may be selected,
which are obtainable with power ratings from ¼W-2W.
The value of the shunt resistance should be selected based on the limitations of the driving amplifier
and the approximate expected value of the load. Selecting too low a resistance here may draw too
much current from the driving amplifier, overheating the resistor or distorting the amplifier. Too high
a resistance may cause the voltage drop across the load to become negligible compared to crosstalk,
calibration errors, etc., producing inaccurate results. In general, the shunt resistance should be
comparable to the expected load resistance and no less than the minimum load impedance for the driving
amplifier, with the calibration resistor in the same range. For example, when testing 4-16Ω (nominal)
loudspeakers with a headphone amplifier, selecting a 50Ω shunt resistor and a 20Ω calibration resistor
is a reasonable choice. However, when testing a 10Ω /70V constant-voltage line, a 500Ω shunt resistor
is probably better suited to the task. Table 2 shows suggested values for the shunt resistor when
testing various loads using both a small power amplifier and a typical headphone amplifier as the
driving source.
Load Rshunt Rshunt

Power Amplifier Headphone Amplifier
4-16Ω loudspeaker 8Ω 50Ω
100Ω /140V line 200Ω 200Ω
10Ω /70V line 500Ω 500Ω
Note: The value of Rshunt should never be less than the suggested load impedance for
the driving amplifier.
Table 2: Suggested shunt resistances for various loads

driven by typical amplifier configurations.
When measuring high impedance loads, the input impedance of the sound card becomes significant to
the measurement, rendering the calculation mathematics ineffective. In general, load impedances
th
greater than 1/10 of the sound card input impedance should not be measured. For loads in this
range, use a quality high-impedance buffer amplifier.
Page 4 of 7
The power dissipation rating for the shunt resistor should be at

least equal to the value calculated by Equation 4. For most Vsrcm ax 2
standard loudspeaker measurements using a headphone amplifier P rating ≥
as the driving source, a ½W resistor is acceptable. More attention

R shunt (Eq. 4)
must be given to this parameter when using a power amplifier. For Power rating for shunt resistors. V
srcmaxis
example, a power amplifier sourcing 2VRMS on its output into a 4Ω the maximum RMS output voltage of the
shunt resistance will require a shunt resistor rating of at least 1W driving amplifier.
(continuous). Following the above guidelines will help ensure that your impedance measurements are
of the maximum accuracy possible, and that the possibility of overheating components or damaging
hardware is minimized.
4. USING THE SMAARTLIVE IMPEDANCE FUNCTION
Building on the preceding review of the Smaart impedance measurement technique and the guidelines
for configuring an impedance measurement circuit, this section will take you through a complete
impedance measurement using SmaartLive. The desired measurement will be the load impedance of a
small 5Ω (nominal) 2-way nearfield loudspeaker.
Step 1: Connect the Measurement Circuit
Based on the recommendations in Section 3, this measurement will be performed using the
headphone amplifier of the computer sound card. Since the measurement is of a low-impedance
loudspeaker, a 50Ω shunt resistor will be used, which is within the load capability of the headphone
amplifier, and is reasonable to optimize the dynamic range of the measurement. Either the single-
ended (Figure 4) or differential (Figure 5) technique may be used to measure the load impedance;
their results are identical assuming the system is correctly calibrated. In either case, the loudspeaker
and calibration resistors are substituted for the unknown impedance Z in the schematics.
Step 2: Configure SmaartLive
The preparation in SmaartLive for running an impedance measurement is similar to that for a transfer
function measurement. The generator must be configured and gains
adjusted to eliminate input clipping. Open SmaartLive and enter
transfer function measurement mode. Turn on the generator, typically
for a synchronized sine-sweep signal. The standard FFT size versus
frequency resolution trade-offs exist as with standard transfer function
measurements. For this example, a 32K-bin FFT is selected, as
measurement time and update speed are not a significant
consideration. Disconnect any load from the circuit, and set the
generator level to provide a signal level close to clipping on the Smaart
measurement input. At no time after connecting a load will a signal
exceed this level (Figure 6a).
If you are utilizing the single-ended measurement technique, you must

match the input gains for the two input channels using the sound card
gain controls, the mixer balance control, etc. If you are using the Figure 6: (a) Adjust signal levels to
differential technique, you may skip this step; the mathematics of the eliminate clipping; (b) Match channel
differential method automatically compensate for differences in levels in single-ended mode.
channel gains. To perform the calibration, disconnect any load from
the circuit and adjust the gain controls until the two channel levels are equal (Figure 6b). This may
also be easily and somewhat more accurately accomplished by looking at the (logarithmic) transfer
function magnitude and adjusting the controls until the curve reaches 0dB across the spectrum.
Page 5 of 7
After performing these adjustments, the impedance mode may be launched

by selecting “Lin” (linear) on the Transfer Function > Amplitude Scale menu
(Figure 7).
Figure 7: Launching
impedance mode.
Step 3: Calibrate with a Reference Resistor
Now that the impedance mode has been launched and the signal levels
adjusted, connect the calibration resistor (discussed in Section 3) in
place of the load. Double-click on the plot area, launching the
calibration dialog box (Figure 8). Select the appropriate circuit topology
(in this case, single-ended) and enter the value of the calibration resistor
(not the shunt resistor) into the “Calibrated Impedance is…” edit box.
In this example, we’ll enter 49.9Ω, which was measured with a precision
ohmmeter. Click OK to finalize the calibration.
Step 4: Perform the Impedance Measurement Figure 8: Impedance calibration

dialog box.
After calibrating the system, the measurement shown on the

display reflects the impedance of the calibration resistor,
which, in this case, is a constant 49.9Ω (Figure 9). Now, the
loudspeaker may be connected as the load, and the actual
unknown load impedance measured. For this example, the
resulting impedance magnitude measurement is shown in
Figure 10.
For this 2-way system, the mechanical resonance of the low-

frequency driver is easily seen at approximately 90Hz, as
well as the high-frequency peak created by the passive
crossover network. One item to note: this loudspeaker is
rated as having 5Ω nominal impedance, but the minimum Figure 9: Measurement of the calibration resistor.
value in the audio bandpass is only 3.9Ω. This information
might influence decisions on how many devices to place
on a single amplifier channel, etc. In more advanced applications, these impedance measurement
results may yield Thiele-Small parameters for single drivers or other design information. This
example demonstrates the accuracy and potential value of the impedance measurement capability in
SIA SmaartLive.
Page 6 of 7
Figure 10: Impedance magnitude measurement of a passive 2-way loudspeaker system.
REFERENCES
L. Beranek: Acoustics.
Woodbury, NY: Acoustical Society of America. 1996.
R. Small: “Direct-Radiator Loudspeaker System Analysis.”

Journal of the Audio Engineering Society, Vol. 20, No. 5, June 1972.
th
J. Irwin, C. Wu: Basic Engineering Circuit Analysis, 6 Edition.
Upper Saddle River, NJ: Prentice Hall Inc. 1998.
A. Oppenheim, A. Willsky, S. Nawab: Signals and Systems.

Upper Saddle River, NJ: Prentice Hall Inc. 1997.
Page 7 of 7
NOTES
SIA-Smaart ®
Schools & Applications Seminars
SmaartLive Keyboard Shortcuts
Presented by:
SmaartLive v5 Keyboard Shortcuts
Analyzer Shortcuts Frequency/Time (x-axis) Range

Operating Mode Zoom Primary In = [Up Arrow]
Impulse Mode = [I] Zoom Primary Out = [Down Arrow]
Spectrum Mode = [S] Move Primary Left = [Left Arrow]
Transfer Function Mode = [T] Move Primary Right = [Right Arrow]
General Controls Zoom Secondary In = [Alt]+[Up Arrow]
Generate Signal = [G] Zoom Secondary Out = [Alt]+[Down Arrow]
Smaart On = [O] Move Secondary Left = [Alt]+[Left Arrow]
Pause = [P] Move Secondary Right = [Alt]+[Right Arrow]
Instantaneous = [Ctrl]+[I] Frequency Zooms (Preset Frequency Ranges)
Auto-Locate Delay (Large) = [L] Frequency (Zoom) Range 1 = [1]
Reseed Average Buffers = [V] Frequency (Zoom) Range 2 = [2]
Load System Preset 1-10 = [Ctrl]+([1]+[10]) Frequency (Zoom) Range 3 = [3]
Save Settings to Preset 1-10 = [Ctrl]+[Shift]+([1]-[10]) Frequency (Zoom) Range 4 = [4]
Print = [Ctrl]+[P] Spectrum Mode Frequency Scale
MIDI Program Change = [Ctrl]+[M] Narrowband = [5]
Decrease Delay Time (0.01 ms) = [F3] 1/24-Octave = [6]
Increase Delay Time (0.01 ms) = [F4] 1/12-Octave = [7]
Clear Delay (Reset to 0 ms) = [F5] 1/6-Octave = [8]
Recall Stored Delay Time Preset = [F6]+[F10] 1/3-Octave = [9]
Spectrum Mode Only Octave = [0]
Trace Difference = [Ctrl]+[F] Trace Shortcuts
Noise Criterion (NC) mode = [Ctrl]+[N] Make Left Input (0) Active = [Shift]+[0]
Reset SPL History Min/Max = [Ctrl]+[R] Make Right Input (1) Active = [Shift]+[1]
Timed Average/LEQ Setup = [F12] Hide/Show Left Input (0) = [Alt]+[0]
Transfer Function Mode Only Hide/Show Right Input (1) = [Alt]+[1]
Phase Display = [F] Hide/Show Transfer Function = [Alt]+[2]
Coherence Function (on/off) = [H] Time Windowed Transfer Function On/Off = [Alt]+[3]
Subtract Reference Trace from Live Trace = [M] Shift (active) Live Trace Up = [Ctrl]+[Up Arrow]
Wrap/Unwrap Phase Display = [U] Shift (active) Live Trace Down = [Ctrl]+[Down Arrow]
Set Phase Range to -180 -> 180 = [Alt]+[Home] Reference Trace
Set Phase Range to 0 -> 360 = [Alt]+[End] Capture to Active Register = [Space Bar]
Swap/Un-Swap Transfer Function Inputs = [W] Select/Show/Hide Reference Bank = [A, B, C, D, or E]
Range, Scale, and Zoom Shortcuts Capture to Selected Register in Bank = [Ctrl]+[A, B,
Quick Zoom = [Ctrl]+[Q] C, D or E]
Amplitude/Magnitude (y-axis) Range Select Next Register in Bank = [Shift]+[A,B, C, D or E]
Zoom Primary In (vertically) = [+/=] Capture to Next Register in Bank = [Ctrl]+[Shift]+[A,
B, C, D or E]
Zoom Primary Out = [–]
Reference Information = [Alt]+[R]
Move Primary Up = [PageUp]
Erase Current Reference Trace = [Ctrl]+[Delete]
Move Primary Down = [PageDown]
Erase All Reference Traces = [Ctrl]+[Shift]+[Delete]
Zoom Secondary In (vertically) = [Alt]+[+/=]
Shift Active Reference Trace Up = [Shift]+[Up Arrow]
Zoom Secondary Out = [Alt]+[–]
Shift Active Reference Trace Down = [Shift]+[Down
Move Secondary Up = [Alt]+[PageUp]
Arrow]
Move Secondary Down = [Alt]+[PageDown]
Save Active Reference Trace = [Ctrl]+[S]
External Device Shortcuts Set/Remove Locked Cursor
Show/Hide Device Bar = [Ctrl]+[V] Set at mouse cursor position = [Ctrl] + Click on plot
External Device Mode = [X] Set at highest peak on the front trace = [Shift]+[P]
Flatten Selected filter = [Del] Set at lowest point on the front trace = [Shift]+[L]
Increase Boost = [Up arrow] Remove Locked Cursor = [Ctrl]+[X] or [Ctrl]+Mouse
Decrease Boost = [Down arrow] click off plot
Increase Frequency = [Right arrow] Move Locked Cursor
Decrease Frequency = [Left arrow] Move to mouse cursor position = [Ctrl]+Click on plot
Increase Bandwidth = [Shift]+[Right arrow] Move to highest peak on the front trace = [Shift]+[P]
Decrease Bandwidth = [Shift]+[Left arrow] Move to lowest point on the front trace = [Shift]+[L]
Select Next Filter = [Tab] Move to next point higher on trace =[Ctrl]+[Shift]+[P]
Select Previous Filter = [Shift]+[Tab] Move to next point lower on trace = [Ctrl]+[Shift]+[L]
External Device Mouse Shortcuts: Track Peak = [Ctrl]+[Shift]+[T]
• Mouse Click on filter marker to select. Move one pixel to left = [Ctrl]+[Left Arrow]
• Click and drag filter marker to change frequency Move one pixel to right = [Ctrl]+[Right Arrow]
and/or boost/cut. Move one data point to left = [Ctrl]+[Shift]+[Left Arrow]
• [Shift]+ Click on plot sets nearest unused filter to Move one data point to right = [Ctrl]+[Shift]+[Right
mouse cursor location or creates new filter at mouse Arrow]
cursor position (depends on device type). Harmonics
Impulse Mode Shortcuts Show Harmonics = [Ctrl]+[H]
Impulse Mode = [I] Next Harmonic = [Shift]+[Right Arrow]
Open Impulse = [Ctrl]+[O] Previous Harmonic = [Shift]+[Right Arrow]
Start/Stop Impulse Recorder = [R] Options Menu Shortcuts
Assign Cursor Position to Delay = [Ctrl]+[Space Bar] Options (All) = [Alt]+[O]
Assign Locked Cursor to Delay Preset = [Ctrl]+[F6- Device Options = [Alt]+[A]
F10] Delay Options = [Alt]+[D]
Impulse Mode Mouse Shortcuts Graph Options = [Alt]+[G] (or Click on Plot Title)
• Click and drag in thumbnail to zoom in on time axis. Input Options = [Alt]+[I]
• Click in left margin of plot to zoom out to full Time Impulse/Locator Options = [Alt]+[L]
scale Preset Options = [Alt]+[P]
Note: In Impulse mode, the Frequency Range Volume (Recording) Control = [Alt]+[P]
commands also function as Time Zoom commands.
External Device Information = [Alt]+[X]
If Locked Cursor is present:
Zoom Options = [Alt]+[Z]
• [Shift]+Click on Impulse mode plot opens Delay
Options, sets Delay Time to Locked Cursor position
If Locked Cursor is not present:
• [Shift]+mouse click on Impulse plot opens Delay
Options, sets Delay Time to mouse cursor position
Cursor Shortcuts
Mouse Cursor
Track Nearest Data Point = [Ctrl]+[T]
Move left one data point = [Ctrl]+[Alt]+[Left Arrow]
Move right one data point = [Ctrl]+[Alt]+[Right Arrow]
NOTES

SmaartLive Class Handout

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SmaartLive Class Handout

Uploaded by

Copyright:

Available Formats

SIA-Smaart ®

Schools & Applications Seminars

All rights reserved

PowerPoint: Class Notes

This PowerPoint presentation can be downloaded at:

Copyright 2004 - SIA Software Company, Inc.

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

SIA Contact Info

Other SIA E-mail Addresses:

SIA Software Company, Inc

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Amp vs Time Amp vs Freq

(*Fractional Octave Banded View)

What do you get if you transform

Frequency Domain to Time Domain

*So . . . If Frequency Response can be measured with music . . .

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

* Decimation – Varying SR & FFT to get constant res.*

The Only Math You Need Today:

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Linear Frequency Scale TC = FFT/SR

Log Frequency Scale

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Linear vs. Log Banding

Linear banding has an increasing number of Fractional–octave (log) banding has

FPPO mode utilizes multiple FFT’s of varying TC to

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Dual-Channel Measurement Issues:

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Given as a value between 0 to 1 (0% - 100%)

100% (Top of Plot)

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

At what SPL should I measure?

Slowly turn up your measurement signal level . . .

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

How Smaart Works

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

How Smaart Works

Input System Output

FFT Spectrum Views

How Smaart Works

Input System Output

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

How Smaart Works

Input System Output

Basic Measurement Set-up

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Class Notes Spectrum Mode

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Spectrum Mode: SPL Calibration

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Three ways to improve the dynamic range of your

3. Increase TC/FFT Size

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

System Alignment Notes

Copyright 2004 SIA Software Company Inc. - a LOUD Technologies Company

Tools in Order of Use

• Solve the problem at the source.