Arbitrary Waveform Generation

with the Verasonics Research Ultrasound Platform

Peter Kaczkowski, PhD

Senior Scientist, Verasonics Inc.

February, 2O16
 Arbitrary Waveform Generation

Overview Table of Contents

The Verasonics Vantage ultrasound system is a Overview.................................................................1
highly programmable platform, designed to enable Introduction.............................................................1
research in array-based sensing, imaging, and high Ultrasound imaging innovation using
energy delivery technology in a wide range of tech- programmable ultrasound platforms.......................2
nical fields. This article specifically addresses the Fundamental limits on image resolution.................2
capabilities of the Vantage transmitter, and how to Conventional transmit waveforms...........................3
use it with a transducer to produce complex analog
Ultrasound bioeffects and safety limits....................3
waveforms. Because of its flexibility, many applica-
Large time-bandwidth waveforms and pulse
tions using various types of modulated waveforms
compression............................................................4
can be explored with the Vantage multi-channel
One-way and two-way impulse responses..............4
arbitrary waveform transmitter.
Pulse coding and compression...............................5
A brief introduction in the use of large Application: Increasing imaging depth...................6
time-bandwidth product waveforms in acoustics is Application: Increasing frame rate..........................6
presented in the context of medical imaging, an Application: Time-reversal methods for reverberant
application which is constrained by safety limits on environments........................................................7
the energy entering the body. A description of the Amplitude Modulation using Pulse Width
Vantage tristate pulser and transmitter hardware, and Modulation............................................................7
waveform software programming interface follows. PWM basics..............................................................7
Basic imaging waveforms can easily be produced, but Converting analog waveforms to PWM waveforms....9
more flexibility is achieved by the use of pulse-width
Vantage pulser constraints.....................................10
modulation. It is important to recognize that when
Programming the Vantage Transmitter..................10
filtered by a transducer’s transfer function, a pulse-
width modulated trinary driving signal can be used to Vantage transmitter hardware characteristics........10
produce analog acoustic waveform over a dynamic Programming the pulser with built-in functions.....11
range that is much larger than might be expected. The Arbitrary Waveform Generator software toolkit...12
Indeed, with knowledge of a transducer’s impulse Application Examples ...........................................13
response, digital pulse trains can be designed to Golay encoding for improved SNR and extended
produce accurate arbitrary analog waveforms: for a imaging depth.......................................................13
5 MHz transducer with 60% bandwidth, coded wave- Transducer compensation......................................16
forms with residual RMS error of nearly -30 dB have APPENDIX A. Frequently Asked Questions..........17
been achieved. A Verasonics software package called
What is the Arbitrary Waveform Generator Toolkit?...17
the “Analog Waveform Design Toolkit” can be used
Why not use a DAC and linear amplifier on every
to determine the necessary Vantage tristate wave-
channel?...............................................................17
forms given a desired acoustic waveform output and
What is the maximum length of an arbitrary waveform
a transducer’s impulse response. Two application ex-
that can be transmitted by the Vantage signal
amples are presented: increased depth of penetration generators?...........................................................17
is observed when using Golay pulse encoding, and
APPENDIX B. References......................................18
axial resolution is improved using a drive waveform
that equalizes the transducer’s spectral response and
effectively shortens its temporal impulse response. 1
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

Introduction Fundamental limits on image resolution

Ultrasound imaging innovation using Ultrasound image quality depends on many fac-
tors, but one of the most important is spatial resolu-
programmable ultrasound platforms
tion. For narrow band signals, image resolution is typ-
Medical ultrasound imaging has undergone
ically limited to scales on the order of half an acoustic
many improvements since its initial development,
wavelength. For acoustic frequency f in a medium of
and clinical systems now produce sophisticated 2D
sound velocity c, the wavelength λ is given by
images in B-mode and Doppler, and moving 3-Di-
mensional volume renderings.
λ = c / f.
Many ideas for improving image quality and
For broadband signals with bandwidth BW, the
information content have been proposed over the
best range resolution ΔR (in the direction of propaga-
years, but have languished in academic and industri-
tion) is
al labs for lack of practical tools with which to imple-
ment new concepts.

ΔR = c / 2 BW.
With the advent of “open” software-based
Acoustic attenuation limits the distance to which
ultrasound systems that provide the developer with
sound can propagate, and the relatively weak back-
access to raw received data, easy programmable
scattering coefficient of tissues further reduces the
control of the transmit-receive sequence and beam
echo magnitude received at the array. The acoustic
parameters, and access to all of the real-time pro-
pressure magnitude decreases exponentially with
cessing functions in the imaging chain, a new era of
propagation distance, and for a plane wave, the
innovation is underway. The software-based ultra-
intensity at distance d is given by
sound system has led to the exploration and devel-
opment of fundamentally new approaches to ultra-

sound imaging including elastographic methods,
I(d) = I0 e -2αfd
new beamforming variants, and the use of coded
where I0 is the intensity on the plane d=0, and α is
pulses. Applications of coded transmission include
the attenuation coefficient, tabulated for various
extending the useful bandwidth of a transducer,
tissues but taken to be about 0.3 – 0.5 dB/(MHz•cm)
extending the imaging depth in attenuating media,
on average. Thus, the maximum practical frequency
increasing the frame rate, and using time-reversal
(corresponding to the best achievable resolution) is
methods to achieve focusing gain that can greatly
the result of a trade off between image resolution
exceed that of conventional array beamforming.
and penetration depth. Given current commercial
The Verasonics research ultrasound platform en-
transducer and amplifier technology, and the lim-
ables exploration and development of all of the
itations placed on acoustic levels by safety limits
techniques listed above. This document specifically
(discussed further below), rules of thumb can be de-
addresses the use of the Verasonics multi-channel
veloped for this trade off, when using short imaging
programmable arbitrary signal transmit capability to
pulses. For example, the ratio between maximum
explore and advance the use of complex waveforms
imaging depth and resolution is approximately
in ultrasound imaging.

Dmax / λ ≈ 400,
2
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

that is, typical acoustic imaging systems can “see” on-screen annotation with the current values of the
into the medium about 400 wavelengths. Because two metrics is still in force today (it is in its third
the sound speed of water is about 1500 m/s, the revision);[2]. A standard for measurement of relevant
acoustic frequency used in medical imaging of soft ultrasonic parameters has also been produced [3].
tissues is on the order of
To prevent formation of bubbles by the field
f ≈ 600 / Dmax (acoustic cavitation), the instantaneous rarefactional
pressure must be held below a prescribed limit, and
with f in MHz, and Dmax in mm.
this places a maximum cap on drive voltage for a
given transmission (defined by the pulse waveform
Conventional transmit waveforms
and beam geometry). The likelihood of cavitation in
Conventional transmit waveforms for B-mode im-
tissue is expressed using a non-dimensional metric
aging are typically short bursts of one or two cycles.
called the Mechanical Index, MI, which is proportion-
Given the physical constraints of propagation and
al to pulse voltage and inversely proportional to the
scattering losses in soft tissues, the image Signal to
square root of the pulse frequency; the MI is scaled
Noise Ratio (SNR) is ultimately limited by safety con-
such that the regulatory limit for the MI in many soft
siderations which restrict the acoustic field’s instan-
tissues is MI < 1.9. (Note that the MI is a single num-
taneous pressure magnitude and average intensity
ber for any given mode and it’s current control setting,
to prevent tissue damage.
defined by the greatest value of the metric computed
for every point in the field.) The maximum allowable
Ultrasound echoes using short pulses are generally
MI thus limits the maximum imaging depth for a given
processed using delay and sum beamforming meth-
acquisition mode using a particular transducer.
ods to reconstruct images, and offer good resolution
(given the center frequency of the transducer’s pass-
The amount of tissue heating resulting from
band) because the transmitted pulses are only slightly
ultrasound absorption is dependent on the average
longer than the transducer’s own impulse response;
power carried by the acoustic field. A non-dimension-
such waveforms are said to have small time-band-
al metric has also been defined to characterize the
width product.
thermal impact (and risk) of a particular acquisition
sequence: the Thermal Index, TI, is the ratio of the
Ultrasound bioeffects and safety limits
average power at a given location in the acoustic field
Safety limits are based on two types of bioeffects
to the power required to raise the temperature in that
caused by ultrasound: mechanical disruption of
location by 1 °C, in steady state and based on specific
tissues due to cavitation, and thermal damage from
tissue thermal models. Thus, the TI must always be
heating produced by absorption of ultrasound [1].
calibrated to the particular acquisition mode (beam
To facilitate regulation and standardization of diag-
geometry, scan format, and PRF), transducer transmit
nostic ultrasound systems, two metrics were devel-
efficiency, and tissue attenuation. The TI value for a
oped by a team of clinical practitioners, researchers,
particular imaging mode is displayed to provide guid-
regulatory and standards officials to provide re-
ance to the sonographer, but the safety limit imposed
al-time indicators of the acoustic regime in a particu-
by the USFDA is expressed in terms of spatial peak
lar imaging mode and for current settings.
acoustic intensity (acoustic power density in mW/cm2),
averaged over the entire imaging frame (spatial peak
The resulting Output Display Standard (ODS) for
3
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

temporal average, ISPTA) or averaged only over the Much of the signal processing formalism dealing
pulse (spatial peak pulse average, ISPPA). In practice, with modulated waveforms was developed in ra-
for scanned focused beams, the TI dependence on dar sensing, and in communications applications
the beam details is rather weak, and thermal damage where the goal is to efficiently transmit information
is in fact related to “thermal dose”, a time-integrated through a complex medium used coded pulses.
quantity dependent on the temperature history. It is Radar imaging uses various modulated methods to
commonly accepted that raising tissue temperature improve the SNR (for improved resolution or extend-
a few degrees for a few minutes is safe; indeed, the ed viewing range), or to reduce the effects of clutter.
WFUMB (World Federation of Ultrasound in Medicine Imaging applications of rapidly moving media, often
and Biology) consensus is that indefinite exposure to encountered in medical ultrasound imaging, use
ultrasound with a TI < 1.5 is safe [4]. coded excitation to increase the frame rate, as will
be discussed in another section.
Details regarding the parameters on which the MI
and TI depend are beyond the scope of this paper Imaging with modulated pulses always requires
(for an excellent exposition with many references, see some form of pulse compression, that is, additional
[1]), but it is important to recognize that, as voltage signal processing of the echo data before high res-
is increased, most acquisition sequences reach one olution images can be produced using conventional
safety limit well before reaching the other, and this reconstruction techniques.
observation provides an opportunity to improve the
SNR by modifying the acquisition parameters such One-way and two-way impulse responses
that both limits are reached at about the same volt- When a voltage drive signal is passed through a
age. For B-mode imaging using short pulses and transducer, the acoustic signal in the acoustic medium
a scanned focused beam, the MI limit is often the is given by a convolution between the drive signal and
dominant constraint, thus permitting the use of longer the transducer’s (transmit) impulse response. Upon
pulses (at similar voltages) until the TI limit is reached. reception, the backscattered acoustic pressure signal
The challenge is then to determine how best to use arriving at the transducer elements is once again con-
longer pulses to produce high resolution images while volved with the transducer’s (receive) transfer function
taking advantage of the increase in acoustic energy to and appears as a voltage on the element before am-
increase imaging depth. plification and digitization. This view leads to one-way
and two-way impulse response definitions. In general,
Large time-bandwidth waveforms transducers operating in their linear range are recipro-
and pulse compression cal in the sense that the impulse response is assumed
Instead of driving the transducer with a short to be the same, regardless of which direction energy
pulse, a longer broadband modulated drive signal is flowing. However, the responses may be different if
can be used, especially in medical B-mode applica- the impedance of the transmitter is different than the
tions that are voltage limited by the Mechanical In- impedance of the receiver, which is true in the case of
dex (MI) and yet have plenty of margin with respect the Verasonics system. The one-way response can be
to the thermal limit (Thermal Index – TI). The modu- measured using a hydrophone, but that estimation is
lation approach depends on the application, and on challenging in water because acoustic propagation
the complexity of the associated time-compression is nonlinear, and the wave becomes distorted be-
algorithms to undo the broadening of the echoes. cause energy from the fundamental drive frequency is

4
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

converted to harmonics. The degree to which nonlin- Pulse coding and compression
earity becomes apparent depends on many factors, The following figure illustrates the use of a large
most importantly signal amplitude and propagation time-bandwidth pulse to deliver more energy per
distance. The two-way response is often sufficient, pulse, and begins with the definition of a transduc-
because imaging applications use backscattered data, er’s impulse response in a linear system. The coded
and therefore the relevant response can usually be pulse is obtained by convolution of the code with
measured directly using the system and transducer, the impulse response. Pulse compression can form
and a well defined target and propagation geometry. an approximate estimate of a system impulse re-
If the waveform in the acoustic medium is important sponse, by selecting a wideband transmit pulse that
to the application, the 1-way transmit impulse re- produces suitably low autocorrelation sidelobes
sponse is needed. (Fig. 1 below). A linear statistical model approach

B. A coded excitation E is a waveform

Impulse Response, Coded Excitation, and Pulse Compression
2 2 with a large time-bandwidth product,
1.5 1.5
1.5
and carries much more energy in the
A
1 1 1
pulse for the same signal magni-
tude (not illustrated because the
∗ =
0.5 0.5

δ (t)∗TD (t) = TD (t)

0.5
BW

curves are normalized). Typically, the

0 0
0
-0.5 -0.5

-1 -1 -0.5
bandwidth of the coded excitation is
2
-1.5 -1.5
0 2 4 6 8 10
similar to the transducer’s bandwidth,
B -2
1 -1 -0.5 0
-2
0.5 -1 1 -0.5 0 0.5 1
1.5
because energy outside the trans-
1 ducer’s band will not be efficiently
300
E(t) 0 0.5 converted to sound.
0
-1 200 C. The code E can be pulse-compressed
-0.5
0 2 4 6 8 10
to produce a short pulse by matched
C
2 -2
-1.5 -1
2
100
-0.5 0 0.5 1 1.5
1.5
filtering; here we use the simplest
1 1
1 choice of filter, and observe that
E(t)!E(t) = δ BW (t) 0 ★
0 0 = 0.5

0
autocorrelation produces a band-
limited impulse. The star operator
-1 -1

-0.5
represents cross correlation.
-2 2 -2
-100 0 2 4 6 8 10
801
D. Excited by a coded pulse, the trans-
-1.5 -1 -0.5 0 0.5 1 1.5
-1.5 -1 -0.5 0 0.5 1.5
2 1.5 1.5
60
D 1
ducer’s output signal S is given by the
-200
40 1
1

∗ -3 -2 20
= -1 0 1 2 3 convolution between the excitation
0.5

E(t)∗TD (t) = S(t) 0 0 0.5

and the transducer’s impulse
0

-0.5 -20
0
-1
-1
-40 response. The resulting waveform is
-2
-1.5 -1 -0.5 0
-1.5
0.5 1 1.5
-60

-80 #10 5
-0.5
0 2 4 6 8 10 longer than TD, and its spectrum is
80 -2 80 -2 -1 0 1 2
1.5 given by the product of the spectra of
E
-1 -0.5 0 0.5 1 1
60 60

40 40 1 E and of TD.

S(t)!S(t) = δˆ(t)
0.5

=
20 20

-20
0

-20
0
★ 0
0.5
E. The acoustic echo response from
0
-40 -40
-0.5 the medium can be modeled as the
-60 -60
-1
-0.5
0 2 4 6 8 10 sum of delayed and scaled replicas
-80
-2 -1 0
-80
1 -2 2 -1 0 -1.5
1 -1 2 -0.5 0 0.5 1 1.5 of the output signal S. To recover the
Figure 1. Impulse Response, Coded Excitation, and Pulse Compression range resolution defined by the signal
bandwidth, matched filtering by cross
correlation with S is applied to the RF
A. A transducer impulse response TD is the temporal waveform resulting from an impulsive excitation;
data. For a single point scatterer, the
the spectrum of that response determines the transducer’s bandwidth, BW. The asterisk operator *
result of such pulse compression is
represents convolution.
another band-limited impulse, with
significant sidelobes.
5
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

can provide far better compression and sidelobe with each other. If different coded waveforms can be
performance [5], [6]. Further complications arise transmitted simultaneously on individual array chan-
when taking into consideration nonlinear phenom- nels, the encoding can be done in such a way as to
ena. For example, attenuation in the medium adds make the individual beam signals distinguishable in
a depth-dependent variation to the pulse shape. post-processing. Thus, many different T/R experi-
Choosing suitable excitations and pulse compres- ments may be conducted in the same acquisition, in
sion methods is an active area of imaging research. the time it takes for one acoustic round trip.

While there are many reasons to use pulse An extreme example of overlapping transmit
coding in Medical Ultrasound and Non-Destructive beams is imaging with plane waves at multiple dif-
Evaluation, the following examples illustrate some ferent angles; the pulses for each angle can be en-
important applications of the approach. Many books coded independently, and the coded and delayed
and articles have been written on the subject; some waveforms superposed into one complex waveform
useful references may be obtained in the articles by for each transmit channel. These emissions ensonify
Misaridis et al. [7]-[9], and in Huang and Li [10]. the entire image field during a single transmit event,
and the resulting echoes can be separated into in-
dividual data sets for each angle in post-processing,
Application: Increasing imaging depth and can then be reconstructed independently as
At a given drive voltage, large time-bandwidth though they were acquired in sequential events. The
pulses contain substantially more total energy than longer the pulse code, the less “crosstalk” remains
conventional pulses. After pulse compression, echo between different codes after processing, and thus
signals appear to have been produced using short the pulse code length is an important design param-
pulses at voltages far greater than the actual drive eter. Note that the independence properties of the
voltage, and also much greater than the maximum coded signals after transduction depend strongly on
permissible voltage given by the MI limit. Thus, in the transducer’s bandwidth and pulse length lim-
their most basic application, coded transmits can be itations. It is the user’s responsibility to choose the
used to extend the maximum imaging depth when coding scheme and post-acquisition compression
the MI limit is more restrictive than the TI limit for algorithm that is most suitable for their application.
short pulses.
One interesting approach developed recently
Application: Increasing frame rate uses Hadamard encoding, where only the polarity
Another common application of coded trans- of the pulse is used to distinguish between trans-
missions is increasing the acquisition frame rate by missions [11]. The goal is to improve the SNR using
performing normally sequential acquisitions simul- conventional (short) pulses, while maintaining the
taneously. For example, a conventional line mode frame rate. By transmitting only two different direc-
imaging sequence performs a series of Transmit-Re- tions at a time, the authors explore combinations
ceive acquisitions, each of which is used to obtain of up to 32 plane waves that are decoded without
one scan line in the image. This procedure is done requiring pulse compression, using very simple pro-
sequentially because the transmit beams for acqui- cessing that merely performs coherent compound-
sitions of adjacent lines overlap and would interfere ing. Another approach to increasing frame rate by
simultaneous coded transmissions on multiple chan-

6
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

nels is disclosed by Flynn et al. in [6] and [12]. The (PWM). PWM can be used very effectively to pro-
method transmits randomly generated binary phase- duce complex analog waveforms given the inevita-
shift keying codes to enable retrospective synthetic ble band pass filtering by the transducer itself.
transmit aperture imaging, and demonstrates addi-
PWM basics
tional benefits of reducing mechanical index while
Digital switching circuits typically transition be-
maintaining sensitivity.
tween two states. Consequently, one might assume
that they would be useless for production of wave-
Application: Time-reversal methods forms with many different voltage levels, that is,
for reverberant environments with high dynamic range. In fact, PWM can be used
Time-reversal applications also require use to perform amplitude modulation with a surprising-
of long time-bandwidth waveforms; in a classic ly high dynamic range, when the digital switching
time-reversal experiment, the coded waveforms are speed is much greater than the highest frequency
measured echoes from transmission of short pulses in the output signal. This approach has become
into a highly reverberant environment, that is, with commonly applied in power audio circuits using
extensive multiple scattering. When these long echo “class D” amplifiers [16], [17], [18], essentially pow-
waveforms are reversed and transmitted back into er switching circuits, to produce acceptable audio
the medium, the propagating waves are naturally quality sound very efficiently. The high efficiency is
pulse compressed by the medium itself, and can a consequence of the rapid transition time between
achieve enormous spatial focusing and temporal on and off states that characterize the devices (often
compression gains [13], [14]. FET transistors); the off state presents a high imped-
ance with little current and power dissipation, and
Amplitude Modulation the on state has low series impedance and allows
using Pulse Width Modulation most of the total power to be dissipated in the load.
In medical ultrasound, advances in electronics Conversely, Class A amplifiers are always operated
and transducer design have permitted increasing within the transition region, and cannot be more
the number of elements in acoustic arrays, and than 50% efficient; they dissipate nearly the same
extending array geometry from 1-D to 2-D patterns. power for all signal output levels, including the
The large channel counts have always forced a “quiet” regime with very low signal output.
compromise between transmit signal flexibility and
cost, with most designs choosing unipolar or bipolar Class A amplifiers were widely used because the sig-
pulsers, with two or three voltage states. Because nal amplification is nearly linear when the output levels
of the significant associated cost in hardware and remain small with respect to the total range of the tran-
power consumption, only a few specialized research sition region. With the advent of much faster devices,
ultrasound systems provide digital-to-analog con- Class D operation has become a practical alternative.
verters (DACs) and linear amplifiers on each channel
(Open System, http://lecoeur-electronique.com The drive pulses can be thought of as a series of
[15]). Therefore most excitation coding is imple- impulses, with magnitude given by the integral over
mented using pulsed signals with only two or three each pulse “on” time. The rectangular drive waveform
voltage levels, and signal design flexibility is greatly is greatly smoothed through low pass filtering by the
increased with the use of Pulse Width Modulation

7
 Arbitrary Waveform Generation

transducer which effectively

integrates over the pulses. The
following figure illustrates the
use of PWM in a classic audio
amplifiers application [18],
where the bandwidth of the
switch is many orders of mag-
nitude greater than the signal
frequency.
Figure 2. PWM encoding in audio signals using Class D amplifiers (Analog Devices).

When the switching transition time is only one or related to the pulse width, defined using the PWM
two orders of magnitude less than the fundamental factor, indicated on the right.
output signal period (as is the case for Vantage),
the pulses may be spaced at one rectangular pulse Note that the amplitude of the fundamental
per half-cycle, and the pulse width is used to de- frequency component of the wave is not exactly
fine the magnitude of the output signal during that proportional to the ratio of pulse on- to off-time
half-cycle. The fundamental frequency of the output (PWM factor, or pulse duty cycle). This is simply a
signal is determined by the spacing between pulses consequence of the Fourier decomposition of a
of the same polarity, as indicated in the diagram in rectangular pulse, graphically illustrated in figure 4.
the next figure below; the amplitude of the wave is

PWM factor
1.0 Amplitude
+V 1.0
0.5 0.7

-V

Period Period

Figure 3. PWM encoding in high frequency signals using a single pulse per half cycle.

8
 a

Arbitrary Waveform Generation

5 -1 -0.5 0 0.5 1 1.5
Duty Cycle
SYSTEM SPECIFICATIONS
 
re   4.   PWM   can   be   used   to   set   the   amplitude   of   a   harmonic   wave,   but   the   mapping   between  
PWM  and
value  aexpressed
nd  m agnitude  ias the integral of the cosine
s  sinusoidal.   Rectangular pulse and first three odd harmonics
xpressed  function overof  the
as  the  integral   the  cinterval defined
osine  function   over  by
the  the on-defined  by  the  on-­‐
interval   1
Duty Cycle = 0.6
time.
!
0.5
𝐴𝐴   =  2 𝑐𝑐𝑐𝑐𝑐𝑐 2𝜋𝜋𝜋𝜋𝜋𝜋 𝑑𝑑𝑑𝑑  
!
𝑠𝑠𝑠𝑠𝑠𝑠 2𝜋𝜋𝜋𝜋𝜋𝜋 0
with          =    
𝜋𝜋𝜋𝜋
with   -0.5

𝑃𝑃𝑃𝑃𝑃𝑃 • 𝑇𝑇
𝑎𝑎 = -1
4 -1.5 -1 -0.5 0 0.5 1 1.5
Duty Cycle
e  following  figure,  the  magnitude  of  each  harmonic  is  plotted  versus  PWM  factor,  
irst,  third,  and  fifth  harmonics.  For  clarity  in  observing  the  nulls,  normalized   Figure 4. PWM can be used to set the amplitude of a harmonic wave, but the
rdIn thethfigure to the right, the magnitude of each mapping between PWM value and magnitude is sinusoidal.
des  of  3  and  5  harmonics  are  presented  in  the  second  figure.  An  example  
harmonic is plotted versus PWM factor, for the
t  frequency  of  5  MHz  is  used  to  identify  the  limited  number  of  possible  PWM  values  
first, third,
e  4  ns  increment   and
size.  N ote  tfifth
hat  tharmonics. For
he  PWM  value   of  0clarity
.5  is  not  inavailable  at  this  
ar  frequency;  observing the
in  general,   nulls,
the   exact  normalized
value  of  PWM  magnitudes
factor  used  in  the  hardware   Harmonic Amplitude vs. Duty Cycle for 5 MHz
1 is  
from  the  rof atio  
3rd between  
and 5th the  harmonics
integral  number  
are opresented
f  4  ns  counts  infor   the  on-­‐time,  divided  
the First
Third
umber  of  counts  defining  a  half  cycle.  Though  the  integral  above  assumes  0.8 symmetry  
Fifth
second figure. An example transmit frequency of
Relative Amplitude

umbers  only),  in  the  Vantage  transmitter  these  counts  are  integer  values.    0.6
5 MHz is used to identify the limited number of
13 0.4
possible PWM values given the 4 ns increment
size. Note that the PWM value of 0.5 is not avail- 0.2

able at this particular frequency; in general, the 0

exact value of PWM factor used in the hardware -0.2
is derived from the ratio between the integral
-0.4
number of 4 ns counts for the on-time, divided 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Duty Cycle
by the number of counts defining a half cycle.
Figure 5. Harmonic amplitudes for a 5 MHz signal. Amplitude of the 1st, 3rd,
Though the integral above assumes symmetry and 5th harmonics, as a function of PWM factor normalized to the funda-
(even numbers only), in the Vantage transmitter mental. The lines represent theoretical functions, and the circles represent
the discrete value of PWM factor available for the 5 MHz wave, given the 25O
these counts are integer values. MHz system clock.

Converting analog waveforms to PWM waveforms audio signals is illustrated in Fig. 2; that method
The sections above have described the use of performs a binary comparison between a sawtooth
various types of waveforms for improving ultrasound waveform and the input analog signal to select the
imaging while satisfying hardware system and clini- state of the switching circuit. The audio speaker, a
cal safety constraints. Nevertheless, the challenge of band-limited load, effectively integrates the binary
creating desired analog waveforms using binary or waveform and filters out the sharp transitions.
trinary pulse sequences still remains.
Another simple approach can be used for the
Conversion of analog waveforms into PWM bi- case in which MHz ultrasound signals are encoded
nary or tristate signals can be achieved in numerous using one pulse per half cycle, the analog signal
ways. The conventional approach used in encoding can be integrated over each half cycle between
9
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

zero-crossings to find the relative magnitude for that quency), but for arbitrary waveforms it is possible to
half-cycle, and this result can be quantized by the exceed the saturation limit, especially when operating
clock interval (4 ns, for the Vantage) to set the pulse at low frequencies (e.g., below 2 MHz) and a series of
width. This rectangular pulse is then aligned with the pulses dominated by one polarity is encountered.
location of the peak in the original signal by adding
ground states before and after the pulse, as needed. To increase the effective number of amplitude
Given a tristate pulser, positive and negative pulses levels that can be encoded, positive and negative
can be treated independently, though it is helpful pulses can be combined: integrating over them can
to maintain symmetry between adjacent positive help overcome the 3 clock minimum dwell time lim-
and negative pulses to limit saturation, as discussed it. Nevertheless, any PWM approach is rather limited
below, and generally to limit low frequency content at higher frequencies (e.g., above 15 MHz), because
that may be detrimental to the circuitry. This simple the 4 ns clock interval becomes a significant fraction
approach may be adequate for a particular applica- of the signal half cycle.
tion, but it neglects the role the transducer plays in
filtering the tristate wave, and is surely less accurate Programming the Vantage Transmitter
than a method that takes the transducer’s impulse
Vantage transmitter hardware characteristics
response into consideration. The Verasonics Arbitrary
The Verasonics Vantage ultrasound transmitter
Waveform Design toolkit includes the transducer’s
uses a tri-state pulser that allows specification of
impulse response in an optimization algorithm that
arbitrary sequences of three voltage levels [+V,0,-V]
produces the tristate waveform that is needed to cre-
at 4 ns clock intervals (250 MHz master clock rate).
ate the desired analog signal. Furthermore, it does so
The voltage V is nominally constant for the duration
while satisfying hardware constraints described next.
of the pulse and is the same for all channels, but
the tri-state sequences can be programmed inde-
Vantage pulser constraints pendently for each channel, providing the flexibility
There are three important constraints for the required for simultaneous independent transmis-
Vantage pulser. First, the 250 MHz master clock sets sions. For example, aperture apodization can be
the width of the smallest interval available to divide achieved by programming different PWM factors to
a pulse: the time grid is discretized by the 4 ns mas- produce different signal magnitudes across an array
ter clock period. Second, the transmit circuit requires of elements. Furthermore, each acquisition event
at most three clock periods to complete a transition (including transmit and receive operations) may be
to any on-state; this “dwell’ time requirement could programmed to emit a unique set of pulse sequenc-
be relaxed to two cycles for voltage swings that are es. The power supply output voltage may also be
substantially less than the maximum 96 V. Third, the changed between events, but the supply is slew-rate
integrated pulse (cumulative sum) must not exceed limited and some transition time must be provided
a limit defined by a transformer saturation threshold. for this to occur.
For the standard frequency Vantage system, the pulse
must maintain a time integral magnitude below 25 The trinary sequences may be of arbitrary length,
V µs, and the maximum length of any pulse is 175 up to a maximum dependent upon the waveform
clock periods (0.7 microseconds). The saturation limit complexity and transmit memory limitations. An in-
is rarely a concern for conventional pulses which have ternal storage format uses a compression algorithm
a high degree of symmetry (except at very low fre- to efficiently represent the trinary sequence and
1O
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

economize transmitter memory usage. This compres-

sion algorithm complicates specification of an abso-
lute maximum pulse length for a complex waveform,
because it is hard to predict how efficiently a particular
waveform can be compressed. Nevertheless, one can
assert that a perfectly incompressible pulse sequence,
that allows for the 3-clock state transitions and uses
all of the transmitter memory, will last more than 150
cycles, or about 30 µs at 5 MHz. The longest duration
achievable in a single pulse is 10 Mcycles long, or 2
seconds at 5 MHz, for a waveform description that fits
within the transmitter memory after compression.

It is useful to observe the result of practical circuit

implementation on the transmitter output waveform
given a tristate drive signal. As mentioned earlier, the
transmitter output transition time between any two
states is about 3 clock cycles due to output band-
width limitations of the transmit pulser. In addition,
the transmit circuitry includes internal impedance
effectively modeled as an R-L series network, with
Figure 6. The tristate drive signal (upper panel) applied to the pulser is
Rs = 8 Ω, and Ls = 1.4 µH. Thus the waveform at the smoothed by the passband of the output FETs (middle panel). The drive
transducer connector combines pulser transitions and signal into a non-reactive 5O Ω load is further modified by the series imped-
ance of the transmitter which has an inductive component (lower panel).
inductive response, as illustrated in Fig. 6 for a reason-
able load impedance, here assumed to be 50 Ω.

Programming the pulser with built-in functions tude and frequency modulated pulse; and a “Pulse
The Vantage software interface provides several Code” specification that represents a direct defini-
different methods to specify a trinary transmitted tion of an arbitrary waveform. All waveform spec-
waveform. The simplest is the “Parametric” spec- ifications are ultimately converted to Pulse Code,
ification for a burst and includes the pulse fre- the Vantage standard waveform format. The Pulse
quency, pulse duty cycle or fractional pulse width, Code is then passed to the compression algorithm,
burst length (number of half cycles), and polarity, and that compressed representation is stored in the
to generate a basic monochromatic pulse. This hardware transmit memory.
specification format is the standard offering, and
the “Arbitrary Waveform” option is required to use The Pulse Code specification contains a se-
additional methods. quence of lines, each of which has five entries [Z1,
P1, Z2, P2, R] representing the number of 4 ns clock
More complex waveforms can be defined using periods for which the transmitter will be zero (Z1
the “Envelope” type, which produces an ampli- and Z2), +V or -V (P1 and P2), and the number of
11
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

times the line is to be repeated, R. The format for the explicit 250 MHz trinary sequence used to drive
the Pulse Code is simply a Matlab (n,5) array, where the transmit pulsers. The process is diagrammed in
n is the number of lines, and is easily converted to figure 7 below.

Functional or Parametric Tristate

Pulse Code
Waveform Definition Sequence

TW.type = 'parametric'; { 0 0 -1 -1 -1 -1 -1 -1 -1 -1 0 … 0 1 1 1 1 1 1 1 1}
TW.Parameters = [Trans.frequency,.67,2,1];

Figure 7. The process of specifying a Vantage transmitter pulse begins with a high level parametric specification of a pulse. This specification is
converted to “Pulse Code”, the intermediate level description of the tristate waveform as a series of cycles, one per line. Z1 and Z2 are clock counts
representing the length of time the waveform is zero, and P1 and P2 are counts for the (+) and (-) states, respectively. The last entry R is an integer
representing the number of times the pulse should be repeated. Finally, the tristate sequence can be easily constructed from the Pulse Code
definition. The tristate waveform is not stored directly in hardware memory: instead, the sequence is compressed to reduce storage requirements,
and decompressed on the fly in the hardware.

The Arbitrary Waveform Generator software toolkit problem that compensates for the transducer’s im-
The pulse programming methods described pulse response.
above allow the user to specify the tristate waveform
that drives the transmit pulser. A user will common- The AWG toolkit (Arbitrary Waveform Gener-
ly specify an imaging pulse using the “Parametric” ator toolkit) has been developed to examine and
description, and accept that the acoustic waveform design the waveforms associated with a particular
in water (1-way waveform) will be substantially dif- transmission. In particular, it is possible to have the
ferent because it will be filtered by the transducer’s tool design a tristate drive waveform to produce (an
1-way impulse response. Similarly, the backscattered approximation of) a desired acoustic waveform, given
result at the receiver will be given by a convolution the transducer’s impulse response. Note that impulse
between the 2-way transducer impulse response, responses provided by Verasonics for particular trans-
the medium scattering transfer function, and the ducer models, and included with the AWG toolkit,
transmitter signal driving the transducer. Depending also include the system transmitter’s impulse response
on the transducer’s transfer function, the received which results in the distortions illustrated in Fig. 6. As
pulse may look very different from the programmed of this writing, only a few transducer responses are
pulse. Indeed, because a resonant transducer im- included with the toolkit, but more will be added over
pulse response is band limited and usually has some time (the user may always provide their own).
“ringdown” or internal reverberation, the result is to
lengthen echo responses from ideal point scatterers. In addition to transducer impulse response
This artifact may be acceptable, but a procedure to compensation, the tool is helpful in producing any
design an excitation pulse that shortens the acoustic desired waveform within the transducer’s passband,
pulse is desirable; such a procedure solves an inverse including time reversal recordings, AM and FM
12
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

encoded signals, waveforms for coded excitation re-

search, and others.

The algorithms used in the toolkit are described

in [5], in the patent application [19] and also in the
presentation slides (titled: Arbitrary Waveforms using
a Tri-state Transmit Pulser) available on the Verason-
ics website. This approach uses knowledge of the
transducer impulse response to develop an optimal
least-squares solution to the inverse problem of find-
ing a trinary pulse sequence that reproduces a given
analog waveform. The residual error for the toolkit
algorithm for a 5 MHz transducer is typically between
-20 and -30 dB spread over the length of the pulse,
depending on the transducer characteristics (low-
er frequencies have more dynamic range in PWM
specification, and hence can produce waveforms with Figure 8. Pulse Width Modulated waveforms used to encode
greater fidelity). impulses of varying amplitude. See the proceedings paper [5] and
a set of presentation slides on the website for details.

The essential idea behind the toolkit’s algorithm

random waveforms. This estimation algorithm and
is to use PWM to encode amplitude variations in the
procedure is not yet included in the toolkit. The toolkit
output signal, as template waveforms (“symbols”).
is a GUI-based program, and its operation is fully
These define a set of magnitude levels that can be
described in a separate manual distributed with the
used to approximate a particular analog waveform
software.
(see Fig. 8). Then an optimization process is used
to produce a close approximation to the desired
acoustic waveform by combining the symbols Application Examples
together, while adhering to pulser constraints de- Golay encoding for improved SNR and extended
scribed earlier. imaging depth
As discussed in an earlier section, large time-band-
The problem of estimating a transducer’s impulse width pulses are often used to improve imaging
response from measurements is not a trivial one. penetration depth when transmit signals are volt-
Conventional methods measure the impulse response age-limited due to hardware constraints or ultrasound
using a broadband pulser and a hydrophone or flat exposure safety limits. As in Fig.1-D, a long excitation
plate reflector. However, many potential sources of is transmitted into the medium, and the backscattered
error can lead to inaccurate results; the accuracy of data must be pulse compressed, Fig.1-E. Cross-cor-
the toolkit-designed waveforms depends on the accu- relation can be used to compress the received data
racy of the impulse response. The paper [5] describes using a matched filter derived from the original exci-
a robust and accurate approach to determining the tation pulse. Typically, the resulting (compressed) data
needed impulse response from a number of trans- suffers from large correlation sidelobes, which add to
mit / receive experiments, using a set of broadband image clutter.
13
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

Complementary pairs of Golay sequences are con- In this example, complementary Golay binary
structed in such a way as to produce correlation side- sequences of length 16 were convolved with a short
lobes that are equal and opposite in polarity, and thus wavelet to produce the tristate drive waveforms. This
two transmit-receive experiments using the two Golay was done by hand: define the pulse code for an
waveforms can be used to cancel the sidelobes in the elementary wavelet at 5 MHz., for both positive and
pulse compression step by simple summation of the negative polarities, and then concatenate the pulse
results. See the work by Nowicki et al. for background code lines following the Golay sequences given by:
information on Golay sequences [20], [21]. Here, the
A = {1 1 1 -1 1 1 -1 1 -1 -1 -1 1 1 1 -1 1}, and
arbitrary waveform synthesis software tool was not
B = {1 1 1 -1 1 1 -1 1 1 1 1 -1 -1 -1 1 -1}.
used to design the pulses. The goal of this exercise
was to demonstrate the use of the Arbitrary Waveform The resulting pulse code produces the transmit-
Generator by directly programming Golay coded ter drive signals illustrated below in Fig. 9, followed
pulse trains using the Pulse Code format to improve by a plot of the simulated acoustic waveforms ob-
penetration depth. tained through convolution of the drive signals with
the ATL L7-4 transducer impulse response.

16-bit complementary Golay pulses

-1 0 1 2 3 4 5 6 7 8 9
Microseconds

Figure 9. Tristate drive signals for 16-bit complementary Golay codes, using a simple wavelet to carry each bit (upper panel), and result-
ing acoustic waveforms for a L7-4 transducer (5MHz, 6O% bandwidth) (lower panel). The panels are registered (the horizontal axes are
correctly scaled and time-aligned) and the lag between electrical drive and acoustic output is clearly observed, as is the lengthening of
the drive pulse due to the characteristics of the transducer’s impulse response.

To show how the complementary Golay pulses that the main lobes are very similar, and the side-
work to reduce correlation sidelobes, the simulated lobes are equal and opposite; when summed, the
pulses are pulse-compressed by autocorrelation to result indeed recovers the L7-4 impulse response.
obtain the transducer’s impulse response alone (sim-
ulating the situation in which there is only one point Two T-R events are required to collect the RF
scatterer in the acoustic medium). This process pro- Data for a Golay pair. Each RF line of echo data
duces the following waveforms, in which it is clear is cross-correlated with the corresponding Golay

14
 Arbitrary Waveform Generation

Simulated Match-Filtered Complementary Golay Pulses

20
Pulse A
10 Pulse B

-10

-20
0 2 4 6 8 10 12 14 16 18

40
Pulse A + Pulse B
20

-20

-40
0 2 4 6 8 10 12 14 16 18
Microsecs
Figure 1O. Autocorrelated signals for each of the acoustic waveforms in Figure 9 (upper panel). The summed result is presented in the lower
panel.

acoustic transmit waveform, and then the two results Finally, an examination of the image line passing
are summed to produce the desired RF data. This through the point targets illustrates how the SNR
pulse-compressed RF data is then processed using improves when using the Golay excitations, at the
the usual data processing chain to form an image. same transmission voltage. In effect, after normaliz-
ing the intensity of the first scatterer, the noise floor
An example set of images comparing imaging appears reduced when using the Golay approach.
performance using the signals described above and
a tissue mimicking string phantom is presented next.

Imaging Example: Complementary Golay Pulses Imaging Example: Complementary Golay Pulses
0.7 dB/cm/MHz Resolution Phantom @ 5 Volts" Results in a resolution phantom"
5 MHz Monopulse ! 16-bit Golay! Transmit!
Burst!

2 cm!
• L7-4 Transducer, 5 Volts!

•Golay RF data match-filtered, !

averaged, and normalized by 16!
!
• All other settings equal (e.g. TGC,
image brightness and contrast)!

• Improved signal to noise results in

deeper imaging range!
Last detectable target!

!
Golay Pulse Encoding Results:!
Last detectable target! • Reduced noise floor!
10 cm! • Improved imaging depth of penetration!
!
Verasonics Inc., Redmond WA, USA 19! Verasonics Inc., Redmond WA, USA 20!

15
 Arbitrary Waveform Generation

Transducer compensation Imaging Example: Transducer Compensation

Gaussian Pulse vs. Monopulse (7-Angle Planewave)
This example uses the Analog Waveform De- "
sign toolkit to improve the temporal response of • Predicted: monopulse vs.
equalized GP (60% BW)"
the transducer, by using a drive signal that partially
• Transmitter pulse sequences"
compensates for the long impulse response. The
• Reconstructed Image Slice
drive signal tends to “equalize” the spectrum of the (7-Angle Planewave)"

transducer, by adding energy to the outer parts of the

passband, and doing so with the proper phase so that
the resulting output signal is shorter than the original
impulse response. Note how complex the new drive
signal is compared to a simple impulse; the resulting
output is indeed more compact, but is emitted later
than the impulse response. When the pulse intensities Verasonics Inc., Redmond WA, USA 21!

are overlapped, the improvement is easily seen.

Example: Transducer Compensation

Monopulse vs. Gaussian Pulse (60% BW) (7-angle PW image)"

Profiled Area!

Monopulse Gaussian Pulse !

5.0 V! 6.2 V!
Verasonics Inc., Redmond WA, USA 22!

16
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

APPENDIX A. Frequently Asked Questions

What is the Arbitrary Waveform Generator Toolkit? What is the maximum length of an arbitrary
The Verasonics Vantage platform has an inde- waveform that can be transmitted by the
pendently programmable transmitter on each chan- Vantage signal generators?
nel that can produce arbitrary tristate waveforms.
The maximum transmit waveform length depends
The hardware can be programmed directly using
on the complexity of the desired waveform and the
high level software to define a tristate drive wave-
output frequency, because a compression scheme
form. The software toolkit provides additional capa-
(based on decomposition of the waveform into nest-
bility to visualize (analog) acoustic signals resulting
ed loops of repeated states) is applied prior to stor-
from the tristate pulse sequence driving a particular
ing the waveform in transmit memory. The simplest
transducer. The kit also includes a tristate waveform
answer is to give an example: at 5 MHz, a completely
design tool which produces a sequence of pulse-
incompressible waveform that fills transmit memory
width modulated pulses that will result in a desired
will last just over 30 microseconds (equivalent of 150
output acoustic waveform, given the transducer’s
cycles). However, because the transmitter bandwidth
impulse response. This is very useful for creating
is limited to about 20 MHz, some repetition of 4 ns
amplitude modulated waveforms, coded excitations,
states will be required and the minimum length of a
and for encoding time-reversal experiment record-
“fully complex” 5 MHz waveform will likely be greater
ings. The Arbitrary Waveform Toolkit is an add-on
than 90 microseconds (>450 cycles @ 5 MHz, equiv-
option that enables the hardware (Arbitrary Wave-
alent). A more accurate estimate requires use of the
form Generators) and includes the additional GUI-
waveform compressor applied to the pulse code rep-
based design software. resentation for a particular waveform of interest. The
pulse length scales inversely with frequency. Note
Why not use a DAC and linear amplifier on that the receivers are active during transmit, but are
every channel? saturated if the transmitter on that channel is active,
Conventional arbitrary waveform generators thus “blanking” the receive during transmission of
uses Digital-to-Analog Converters (DAC) and lin- the pulse.
ear amplifiers, and these are straightforward to
use and provide high dynamic range drive signals,
but very expensive to implement in large channel
count systems, particularly when these require high
power, and are far less electrically efficient as well.
Unless extremely accurate reproduction of specific
waveforms is required, using a tristate pulser with
PWM control on each channel is a much more cost
effective hardware solution. Furthermore, Verasonics
provides the software to design the tristate pulse
train given a desired analog waveform as part of the
Arbitrary Waveform Generator Toolkit.

17
 Arbitrary Waveform Generation
SYSTEM SPECIFICATIONS

APPENDIX B. References

[1] T. L. Szabo, Diagnostic Ultrasound Imaging: Inside Out, 2nd ed. Academic Press, 2O14.
[2] AIUM-NEMA, “NEMA UD 3 - STANDARD FOR REAL-TIME DISPLAY OF THERMAL AND MECHANICAL ACOUSTIC
OUTPUT INDICES ON DIAGNOSTIC ULTRASOUND EQUIPMENT - REVISION 2,” NEMA, Jun. 2OO4.
[3] AIUM-NEMA, “NEMA UD 2 – ACOUSTIC OUTPUT MEASUREMENT STANDARD FOR DIAGNOSTIC ULTRASOUND
EQUIPMENT – REVISION 3,” NEMA, Rosslyn, Virginia 222O9, UD 2-2OO4, Jun. 2OO4.
[4] S. B. Barnett, “WFUMB Symposium on Safety of Ultrasound in Medicine. Conclusions and recommendations on thermal
and non-thermal mechanisms for biological effects of ultrasound. Kloster-Banz, Germany. 14-19 April, 1996. World
Federation for Ultrasound in Medicine and Biology.,” Ultrasound Med Biol, vol. 24, pp. i–xvi– S1–58, 1998.
[5] J. A. Flynn, P. J. Kaczkowski, K. Linkhart, and R. E. Daigle, “Arbitrary waveforms using a tri-state transmit pulser,” presented at
the Ultrasonics Symposium (IUS), 2O13 IEEE International, 2O13, pp. 41–44.
[6] J. A. Flynn, L. Pflugrath, P. J. Kaczkowski, and R. E. Daigle, “Coded excitation reconstruction by impulse response estimation
and retrospective acquisition,” presented at the Ultrasonics Symposium (IUS), 2O15 IEEE International, 2O15, pp. 1–6.
[7] T. Misaridis and J. A. Jensen, “Use of modulated excitation signals in medical ultrasound. Part I: basic concepts and
expected benefits,” UFFC, vol. 52, no. 2, pp. 177–191, 2OO5.
[8] T. Misaridis and J. A. Jensen, “Use of modulated excitation signals in medical ultrasound. Part II: design and performance
for medical imaging applications,” UFFC, vol. 52, no. 2, pp. 192–2O7, 2OO5.
[9] T. Misaridis and J. A. Jensen, “Use of modulated excitation signals in medical ultrasound. Part III: high frame rate
imaging,” UFFC, vol. 52, no. 2, pp. 2O8–219, 2OO5.
[10] S.-W. Huang and P.-C. Li, “Arbitrary waveform coded excitation using bipolar square wave pulsers in medical ultrasound,”
UFFC, vol. 53, no. 1, pp. 1O6–116, 2OO6.
[11] E. Tiran, T. Deffieux, M. Correia, D. Maresca, B.-F. Osmanski, L.-A. Sieu, A. Bergel, I. Cohen, M. Pernot, and M. Tanter,
“Multiplane wave imaging increases signal-to-noise ratio in ultrafast ultrasound imaging.,” Phys Med Biol, vol. 6O, no. 21,
pp. 8549–8566, Nov. 2O15.
[12] J. Flynn, R. E. Daigle, and L. Pflugrath, “Method and System for Coded Excitation Imaging by Impulse Response
Estimation and Retrospective Acquisition.” US Provisional Patent Application No: 62/141,749
[13] B. E. Anderson, M. Griffa, C. Larmat, T. J. Ulrich, and P. A. Johnson, “Time Reversal,” Acoustics Today, vol. 4, no. 1, pp.
5–16, 2OO8.
[14] M. Fink and C. Prada, “Acoustic time-reversal mirrors,” Inverse Probl, vol. 17, no. 1, pp. R1–R38, 2OO1.
[15] J. A. Jensen, O. Holm, L. J. Jerisen, H. Bendsen, S. I. Nikolov, B. G. Tomov, P. Munk, M. Hansen, K. Salomonsen, J. Hansen,
K. Gormsen, H. M. Pedersen, and K. L. Gammelmark, “Ultrasound research scanner for real-time synthetic aperture data
acquisition,” IEEE UFFC, vol. 52, no. 5, pp. 881–891, May 2OO5.
[16] MAXIM-DALLAS, “Class D Amplifiers: Fundamentals of Operation and Recent Developments - AN3977,” Mar. 2OO7.
[17] J. Honda and J. Adams, “Class D Audio Amplifier Basics - AN1O71,” International Rectifier, pp. 1–14, Feb. 2OO5.
[18] E. Gaalaas, “Class D audio amplifiers: What, why, and how,” Analog Dialogue, 2OO6.
[19] J. A. Flynn, P. J. Kaczkowski, B. J. Pflugrath, and L. S. Pflugrath, “Method and system for arbitrary waveform generation
using a tri-state transmit pulser.”, USPTO WO 2O15/OO996O A1.
[20] A. Nowicki, M. Lewandowski, W. Secomski, J. Litniewski, and R. Tymkiewicz, “High frequency imaging using coded golay
transmission,” IUS 2OO5, vol. 4, pp. 2O58–2O61, 2OO5.
[21] A. Nowicki, I. TROTS, W. Secomski, and J. Litniewski, “Golay’s codes sequences in ultrasonography,” Hydroacoustics
Selected full texts, vol. 7, pp. 173–182, 2OO4.
18