You are on page 1of 10

THE EFFECT OF SOUND VELOCITY ERRORS ON

MULTI-BEAM SONAR DEPTH ACCURACY

Donald F. Dinn Bosko D. Loncarevic Gerard Costello


Engineering and Technical Services Geoscience Atlantic Canadian Hydrographic Servik
Dept. of Fisheries and Oceans Dept. of Natural Resources Dept. of Fisheries and Oceans

Bedford Institute of Oceanography, P.O.Box 1006, Dartmouth, N.S., Canada, B3A 3V3

-
Abstract A number of hydrographic organizations the position of the ship, or more precisely, the
and commercial operators are using multi-beam sonar position of some point on the ship, e.g., the GPS
systems for mapping shallow and deep water areas. antenna;
With transducer motion and refraction playing an the pitch, roll and heading angles of the ship
important role in measurement accuracy, errors in ship relative to the above position;
motion data and sound velocity profile information are the vertical position of the transducer with respect
major contributors to inaccuracies in the measured to the average water level, i.e., the ship's draft and
depth. The effect of errors in sound velocity is heave at the transducer, (increasingly, with precise,
examined is shown to be different for flat-level high-rate GPS position fixing in three dimensions,
transducers and for tilted transducers or those with the earth ellipsoid is becoming the reference
sector steering. On survey vessels where roll is surface for vertical control);
significant, the roll can modulate the depth errors the changes in water level due to tides and
contributed by sound velocity uncertainty. These errors atmospheric effects;
are discussed in relation to the IHO standards for the profile of sound velocity vs. depth, a basic
surveys, and methods of error control are suggested. determinant of the path of the acoustic ray;
the vector distance (x, y, z ) in the ship frame of
reference between the transducer and the
1. INTRODUCTION positioning sensor (the GPS antenna).

Multi-beam sonars measure the two-way travel time of By virtue of their error propagation characteristics, roll
a short acoustic pulse or tone burst as it travels from angle, sound velocity profile (SVP), and vertical
the transducer to a point (or area) on the sea floor and control are the more significant parameters affecting
back along a path whose geometry is dependent on a depth accuracy.
number of factors. Typically, some 60 separate paths
spanning 150 are "sounded" across the ship's track on
O Much work has been sponsored in recent years by the
each acoustic ping. With the improvements in acoustic Canadian Hydrographic Service (CHS) and others
transducer design and digital signal processing in (Cervenka et al, Dinn and Loncarevic, Dodds, Hughes
recent years, multi-beam echo sounders have become Clarke and Godin, Lachapelle and Loncarevic,
cost effective tools and are being increasingly Loncarevic (1993), and Luscombe) to understand the
employed in ocean mapping, route surveys and measurement errors in the above parameters and their
channel surveys. In addition, the calibrated target- effect on the accuracy of the final geo-referenced
strength information available from some multi-beam depths. In particular it has been shown (Loncarevic
sonars makes them valuable in surficial geology and Scherzinger) that in a realistic error budget for the
mapping and bottom classification. EM1000 multi-beam sonar (manufactured by Simrad
AS, Horten, Norway), the rms uncertainty in roll angle
In the common case where the transducer is not p should not exceed 0.0Y if the uncertainty in depths
mechanically leveled, +e travel-time measurement is measured by the outer beams (ray angle B of f75 ") is
made in the frame of reference of the ship on which to be within &I%, 90% of the time. (The percentage
the transducer is mounted. A number of parameters, error in depth due to roll uncertainty is given by Ad =
many of which are measured at the time of the ping, ApTan(B), where angles are in radians). The
are required to transform the measured slant-range International Hydrographic Organization (IHO)
travel times into accurate geo-referenced depths (x, y, specifies a 1% error limit for chart depths greater than
z; latitude, longitude, depth). These parameters 30m and a 0.3m error limit for depths of 30m or less,
include: with a 90% confidence limit.

1001

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
Extensive testing and recent survey work carried out time for the taking OF SVPs. Under-sampling the
by CHS and the University of New Brunswick on sound velocity in the survey arca - either spatially or
Stellwagen Bank off Massachusetts using the survey temporally - can result in dcpth data of poor quality
ship Frederick G. Creed has demonstrated that a new To avoid slowing the smooth progress of the survcy, it
GPS-aided inertial sensor, POS-MV (manufactured by may be necessary to use equipment capable of
Applied Analytics Corporation: Markham, Ontario, profiling while underway. CHS is currently
Canada) now in use by CHS on the Creed, achieves examining the merits of this approach..
the required accuracy in roll regardless of the extent or
severity of ship motion or maneuvers. Manufacturers of multi-beam sonars do not fully
present, in a quantitative way, the implications of
With the roll error under Control, CHS has turned its spatially under-sampling the sound velocity. The
attention to the two remaining, significant sources of work described here attempts to present the effects of
error: refraction and vertical control. This paper these errors in a way that helps the hydrographer make
examines refraction effects. rational decisions about the need to take a new SVP.

2. REFRACTION-INDUCED ERRORS 2.1 Pointing-Angle Errors

Refraction-induced errors in measured depths show up By using a simplified W o r m e r of two elements for
in digital terrain models of the sea floor as ridges or as analysis (Fig. l), the launch or pointing angle U of a
furrows at extreme beam angles (typically where the beam can be determined using basic physics. The two-
edges of adjacent swaths abut or overlap). Ridges or element approach simplifies the analysis and has the same
furrows occur depending on whether the actual pointing angle and sensitivity to sound velocity as a multi-
refraction effect is greater or less than that determined element transducer; only the beam width and directivity
by the measured SVP. However, as will be shown, the gain are diEerent. M e r a ray is launched from the
instantaneous roll angle modulates the refraction effect transducer, its path is controlled by the SVP <and Sncll's
in a way that is determined by thc type of transducer constant k; k = Sn(B)/c and c is the speed of sound.
used.
The beam pointing angle 0 from vertical, or a From
Given the launch angIe of an acoustic ray, the path of the central axis of the transducer, is a function of the
each ray is determined by the physics of refraction actual sound velocity coo and the measured or
(Snell's Law) and the SVP. Thus, there are two estimated sound velocity cmo at the face of the
aspects to the refraction problem: transducer (in the area AOB in Fig. 1). During a
w e y , values of cm0 are determined by direct
0 the error in the launch angle of the acoustic beam measurement ,using a "sing-around" sensor, or are
due to poor knowledge of the sound velocity at the I -
estimated from oceanographic conditions primarily
transducer €ace, and
the ray path deviation from the calculated path due
to errors in the SVP.

Instrumental errors of sound velocity profilers


(typically i O . 3 d s ) can be i n s i w c a n t in comparison
to the eflects of under-sampling the sound velocity in
areas where it is subject to moderate temporal and
spatial variability. This is particularly true in the
surface layer where solar heating, tidal mixing and
salinity changes (from fresh water input) can be
significant.

Hydrographers face competing demands when


carrying out a multi-beam survey. On one hand, they
must make cffective use of cxpensive ship time to Fig. 1. A simplified two-clement beamformer in which 0
gather sounding data. On the other, they must ensure is the pointing angle, pis the roll angle, j3 is the mounting
the data quality is within standards; this may mean or sector angle and a is the angle from the central axis of the
reducing area coverage ratcs in order to allocate more transducer sector to the steered beam; a = 0- p p . -
1002

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
temperature t and salinity s. A number of empirical By substituting the expression for Eqn. 1 into Eqn. 3, the
relationships exist for c as a function of t, s and actual beam steering angle can be shown to be:
pressure, p or depth d (Burdic, 1984). These
relationships show typical sound velocity sensitivities
of 4.6m/sl0C, 1.3ddppt of salinity and 0.016ddm of
depth. Using the value cfflo, the beamformer
determines the phase delay # required for each where 0, is the value of &?om Eqn. 3 when ,c is used
element in the array in order to steer the beam at the in place of C,O. Note that when p = p= 0, that is, when 0
desired angle (Fig. 1). Clearly, changes in surface = a,Eqn. 4 is fundamentally Snell's law for reftaction
temperature and salinity can have a critical bearing on (see Eqn. 7). This is not true when p + P;c 0, i.e., when
phase controlled beamforming. the transducer is rotated mechanically or electronically.
Thus, when B = a, the pointing angle error has the same
Design values of phase shift 4 d for each element, magnitude and sign as the refraction angle that would
calculated for a given sound velocity cmonear the face result if the ray travelled through a boundary from water
of the transducer, are normally stored in the with sound velocity,c into water of velocity c,. It's as if
beamformer processor as Delay A and Delay B. Once the desired or calculated pointing angle @d is established
calculated, these values are constant until a new sound in a fictitious water layer with sound velocity ,c after
velocity is input. In Fig. 1, the acoustic ray arriving at which the ray passes into the actual water and takes up an
element A of the transducer experiences an actual angle 0,. Eqn. (4) has some intuitive aspects: neglecting
phase lag 4a relative to rays arriving at element B. roll, if the mounting angle is the same as the beam
This phase lag is the acoustic travel time through the pointing angle, q4 will be zero; i.e., no phase-controlled
water over the distance OA. At the design pointing beam-steering is needed. Equally, for an installed-level
angle under conditions of accurately known sound transducer9if the roll angle equals the desired pointing
-
velocity, #d qja will be zero for every element in an angle, #willagain be zero.
array. This is equivalent to creating a virtual broad
side array aimed at angle B. Note that while +!+ isi The sensitivity of the beam pointing angle to
fixed by the value of cffloinput to the processor, is variations in the speed of sound near the transducer
determined by the true sound velocity in the water near face can be found by differentiating a rearranged Eqn.
the transducer, i.e., in the area O m . If c,, differs 3 with a = B - p - flusing the standard form for Sin-'
from cm0 there will be errors in the beam pointing followed by some manipulation. To wit:
angle, because @d will not equal 45,. This leads to:

'In 0

where
a = S-p-p, (2)
p is the roll angle, ,8 is the mounting angle of the array Therefore, as p and p can be considered fixed,
measured to the central axis of the transducer or to the
center of the active sector of a circular transducer, d is A B = -Tan(B
''a 0 - p - p> (6)
the inter-element spacing, and j is the acoustic 'a 0
acoustic carrier frequency. Angles are defined such
that down is 0' and a horizontal line toward starboard where Acu0 represents the uncertainty in c , ~ .Eqn. 6
is +90°. For the lowest side-lobe ratio and the indicates that the beam pointing angle error AB
maximum angular ambiguity resolution in the formed (radians) is a function of both the error in sound
beam, the value of d is usually selected as one-half the velocity dcaonear the transducer, and the amount of
acoustic wavelength, i.e., U2 (Burdic, 1984). phase-controlled beam steering needed. For the case
of a flat, level transducer, a pointing angle error of
Using a value of 4 (bC)calculated from Eqn. 1, and 0.014' occurs when AC,O = I d s and the desired
the actual value of c (ca0) in the water, yields'an launch angle is 75'.
expression for the actual launch angle 0, from Eqns. 1
and 2. The Simrad EM1000 transducer uses a semi-circular
array of elements spaced 1.25"; a 40" sector
containing 31 contiguous elements is selected in
synchronism with the roll angle such that the central

1003

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
Table 1. Pointing Angle Errors for Various Multi- The Elac BCC system is similar to the two transducer
beam Sonars at Zero Roll Angle and option for the EM3000 except that the two arrays are
lmls Error in Sound Velocity normally installed at +30° and -30".

System Transducer
1
I
Mounting,
Sector
Angle, p
1
I
Pointing
Errorfor
=Ids
Table 1 displays the pointing angle error of a number
of multi-beam sonar systems for a I d s sound velocity
error. The results in the table cannot be considered alone,
however, as the pointing angle error is only one part of
Simrad the refracton problem - it is tightly linked to the
EMlOOO refraction effect at the point the ray leaves the area AOB
in Fig. 1.

2.2 Ray-Path Errors


Simrad Flat, level 0" 0.06" for
EM3000 16,= 60"
At each layer 0 to n (where 0 represents the transducer) in
0.143" for a horizontaly-stratified ocean in which the SW is known
14 = 75" with adequate detail, the ray angle 8 through the water
Simrad
EM3000 tilted 45" 1 45" 1 0.02" for
14=750
can be determined using Snell's Law (k is Snell's
constant):

Reson Circular, 0" 0.00" for


Sea Bat Sector- all angles
Steered
The subscript a denoting actual can be omitted from the
ELAC Flat, dual 30" 0.02O for general term in Eqn. 7 because Snell's constant operates
BCC tilted 30" (e( = 60" at all layers, whether the value of c is actual or measured
0.038" for (witherrors).
161= 75"
The specialcase of a flat, level transducer (i.e., p = p= 0)
leads to an interesting result from Eqns. 4 and 7: the
axis of the sector is within 0.62" of the desired deviation in Ou0 created by an uncertainty in coois totally
pointing angle. In this way, a minimum amount - compensated when the ray reaches water where the sound
only i0.62"- of phase-controlled beam-forming is velocity is accurate, i.e., where=,c c., In such a case
needed to create a stabilizedbeam up to angles of 60 O - the actual and the calculated ray angles agree: 8, = 0,.
p (where p is the roll angle), at which point the Note, however, that the ray propagation errors
transducer runs out of elements on the high side. Thus, accumulated up to the layer where c, = ,c,, still
when roll is zero, the EMlOOO has an inconsequential contribute to the overall depth error. Nevertheless, for
-
pointing angle error of Tan(0.62")/1500 radians or errors related to uncertainties in the surface sound
0.0004" up to 0 = *60° for a I d s error in ca0. When velocity, a flat, level transducer has advantages over other
8 = *75 ", the error increases to Tan(75" - 60 ")/1500 = arrangements. The advantage is reduced in rough
0.01" weather when the m e y vessel can roll some *loo, as
will be shown.
The Reson Sea Bat also has a circular transducer, but
does not stabilize the formed beam for roll. Thus, By applying Eqns. 4 and 7 to successive layers, a ray of
while the beam moves with the roll of the vessel, the known travel time (the basic measurement made by the
pointing angle error due to dcn0 of I d s is incon- sonar) can be traced through each layer to the ocean
sequential. bottom. TheQ by interpolating the travel time in the final
layer above the bottom, depth and offset can be derived
The EM3000 employs a flat transducer that is from the path length and angles. The depth values that
normally installed level. As an option for wider swath result from two different sound-velocity profiles (SW),
coverage, two transducers can be installed at angles of one taken as actual, the second in error by some amounc
45O and -45O. The beam is not stabilized to canthen be compared to arrive at the depth error.
compensate for roll.

1004

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
0 EM3000 110.95 EM3000 1 13.09
ry
0.1 43"
I
0.26m

n" 111.17
.... .
iy

0.20m

Fig.2. Simplified ray diagrams (not to scale) for three different bcamforming transducers showing the depth error &, and
lateral position error Ay in 3Om of water. "T" indicates the true bottom;"M"indicates the measured bottom. For a), b) and
c), the actual sound velocity is assumed to be 150lm/s in the surface layer (0 to 0.3m) and 150Ods elsewhere; the mcasured
value is iso-velocity (15oomls) over the full water column. For d), e) and f) the actual sound velocity is 1501ds over the whole
water column; the measurcd protile is iso-velocity (15OOmls). The symbol in each diagram shows tho vertical orientation of
the two element array and the amount of phase-controlled beam steering employed, as measured from the ccntral axis
of the array. The single-transducer Simrad EM3000 and the Elac BCC do not normally operate beyond S O ' ; the ray
plots at 75" arc prcscnted for comparison of generic systems at extreme anglcs.

Using the above approach, ray paths were calculated for a values of 1495ds to lSOSm/s, tapering to 1 5 O O d s at
number of transducers when a I d s error exists in a thin 30% of depth (seeFig. 3). These kinds of profiles are not
(0.3m)surface layer at the face of the transducer. Water atypical: the deeper part of the water column is normally
depth was assumed to be 30m. Results are shown in Fig. more stable than the top layer which is affected by solar
2 a), b) and c). For comparativepurposes, Fig. 2 d), e)and heating, tidal nixing and freshwater influx. The
illustrate ray paths arising from an o B t of I d s over memured SVP was assumadto be iso-velocity at 150Ods.
the whole SVP. This kind of error may not be unusual; Using Eqns. 4 and 7,and the approach dcscribcd above,
an SVP instnrment chat is out of calibration by only the expected depth errors for a number of multi-bcam
0.07% will give approximately I d s error over the systems were calculated as a function of surface sound
normally encountered range of sound velocities. In this vdceity error and roll angle. The results are presented in
case, the flat, level array performs poorly because the Fig. 4 and are not at dl intuitive. In &ct, one of the
pointing angle error persists over the whole my. Table 2 problems in looking at SVP-indud depth errors is to
summarir's the depth errors for various transducer find a simplistic way of pmenting the cause and c&ct,
arrangements. Note that not all of the systems can steer when the error process involves integration over thc
h s to Et75O. whole water column.

For a detailed look at dcplh errors, a number of SVP's The most interesting feature of Fig. 4 is thc difference
were assumed to represent actual conditions with mrfhce created by electronic rotation (sector selection) in the

1005

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
1.o 1.O 1.o

0- 0-
E*0.5 0.5 E- 0.5
2
B
[I:
2 2tY
w 0.0 % 0.0 0.0
I I I
I- I-
Q
W
k n
LLI
a0.5 E-0.5 ao.5

-1 0 -1. .-
0 -1 0
0 25 50 75 0 25 50 75 0 25 50 75
RAY ANGLE, degrees RAY ANGLE degrees RAY ANGLE degrees

1.o 1.o

0-
z- 0.5 f 0.5
2
IY
20:
w 0.0 w 0.0
L I
l
i
w kW
0 -0.5 0-0.5

-1 .o -1.o
-75 -50 -25 0 25 50 75 -75 -50 -25 0 25 50 75 -75 -50 -25 0 25 50 75
RAY ANGLE, degrees RAY ANGLE, degrees
RAY ANGLE, degrees

(4
1.o 1 .o 1.o

a 0.5
i
z 0.5
0-
2 0.5
2LI: 2w s
LI:
w 0.0 w 0.0 I0.0
I I
t
w
t
W w
0 -0.5 D -0.5 ao.5

-1 .o -1 .o -1.o
0 25 50 75 0 25 50 75 0 25 50 75
RAY ANGLE, degrees RAY ANGLE, degrees RAY ANGLE,degrees

Fig. 4. Depth errors for several multi-beam sonars as a function of ray angle from vertical, roll angle, and various
sound velocity profiles with surface differences of -+lm/s,*2.5m/s, and compared to an assumed iso-velocitywater
column. Fig. 3 shows the "actual" S W s . Graphs a), b), c), g), h), and i) are plotted for 0 to 75" ray angles because
each half of the transducer handles opposite quadrants and the errors are symmetrical for positive and negative ray
angles. In d), e), and f) a single transducer covers two quadrants. The single-transducer Simrad EM3000 and the Elac
BCC do not normally measure beyond S O o ; the graphs are plotted to 575" for comparisonof generic systems.

EM1000, by the fixed mechanical rotation (30') in the types of transducer arrangements, plots of depth error
BCC, and by the lack of rotation in the EM3000. vs. surface sound velocity (for zero roll angle) and
depth error vs. roll angle (for a I d s error in surface
To further characterize the error performance of three sound velocity) are shown in Fig. 5 and Fig. 6 .

1006

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
Table 2: Depth Errors (YO)for Various
Multi-beam Systems at Zero Roll Angle
sv ps

Circular
EM3000 0.00 -0.01 0.13 0.86
Flat 0 10 20 30
EM3000, 0.17 0.77 0.04 0.08 DEPTH, metres
flat, dual
SWs used in analyzing depth errors for various
45"
multi-beam sonars. Labels on profiles indicate the dif-
Sea Bat 0.20 0.91 -0.07 -0.08 ference in the surface sound velocity compared to 1500m/s.
dual
circular means that further mechanical rotation due to roll angles
Elac BCC 0.13 0.67 0.00 0.18 of up to 10" does not appreciably affect the error
flat dual performance of the extreme beams. The minor
300 differences in Fig. 4 a), b), and c) are discernible on close
inspection.

3. INTERPRETATION The BCC, with a fixed 30' mounting attitude, has less
depth error for a given ray angle than the EM1000, but it
Although the MO standard for the accuracy of charted exhibits somewhat more variation in error with roll.
depths is 1% for depths greater than 30m, in a realistic
scenario only 0.3% to 0.5% can be assigned to rdi.acton The EM3000 at 0" mounting angle exhibits the lowest
and pointing angle effects. The remainder of the error error up to beam angles of *75" at zero roll angle. The
budget is statistically used up by errors from other sources sign of the error for extreme angles is opposite that ofthe
such as the motion sensor and the bottom detection EMlOOO and the BCC (refer to Fig. 2). The error also
hardware and algorithm. Typically, then, the rms depth shows an interesting change as a function of roll angle.
error due to sound velocity effects should not exceed about At p = Oo, the contributions to depth error from average
0.5% or 0.15m in 30m of water. sound speed and from refraction cancel each other at 8 =
45". This feature has sparked the suggestion @odds,
While it is not shown explicitly in Figs.2 and 4,it should 1994) that through appropriate survey design the 45'
be obvious that the actual depth has no bearing on the beam from one swath be compared to the 0' beam fiom
percentage error in the measured depth; only the an adjacent swath to yield two fairly accurate,
normalized shape of the SVP is important. Thus, a 1 m/s independent measures of depth. Differences could be
error in a surface layer 3m thick in 300m of water gives used to adaptively fine-tune or interpolate sparse SVP data
the same percentage depth error as a 1 m / s error in a until both measures agree. Fig. 3 d), e) and f) indicate
layer 0.3m thick in 30m of water. Note that in water less that in conditions of high roll, the ray angle at which the
than 30m deep, the IHO error limit of 0.3 metres allows errors from average sound speed and those from refiraction
the percentage error to grow as the water becomes cancel each other does not remain constant. Statistically,
shallower. however, the errors should be lowest at ray angles near
45".
Fig. 4 shows, as expected, that all systems have the same
level of error at zero beam angle. In this case, there is no For the EMlOOO it is clear from Fig. 4 that care has to be
refraction to contend with and the error is a function only taken to determine the surface sound velocity to -0.7m/s,
of the average sound velocity over the ray path. Systems ifthe error contribution is to be held below 0.5% at k75'.
like the EMlOOO and the SeaBat with circular transducers It can also be seen from Fig. 4 that the EM3000, with its
have negligible pointing error; practically all of their error better tolerance of surface velocity error, needs a sound
is due to refraction effects. The large amount (WO") of velocity accufacy of only 3 S d s to achieve 0.5% depth
sector steering (electronic rotation) used in the EMlOOO accuracy at 6 0 ' beam angle under conditions of -10"

1007

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
roll. If the sound velocity were known to iO.7m/s, the 2.0 h I 1
EM3000 could bc uscd at beam angles up to &70°
combincd with roll angles up to *loo. 5 1.5
E
The information presented in Fig. 5 highlights the 91. 1.0
rY-
dramatic diEerence between level and sector-steered (or
tilted) sonar arrays when there are errors in the surface
s 0.5
t
Y
W
sound velocity. The graph for the EM1000 explains the E 0.0
depth errors Seen during the testing OF motion sensors for U
I I I
-0.5
multi-beam sonars on the CHS survey vessel Frederick G. I
1 I
BCC t15"
I
Creed The vessel was rolled 4 5 " at a 6-7 second period -1 .o
using its roll-stabilizer fins while under-way at 10 kts. -10
+I"

-5 0 5 10
ROLL ANGLE, degrees
The Creed, shown bow-on in Fig.7, is a 20m twin-hull
vessel with the sonar transduwr mounted in the starboard Fig. 6. Depth error vs. roll angle for cxtrcmc bcam
sponson. Consequently, the transducer moves vertically angles for various multi-bcam systems when tho surface
layer (0.3m) has an actual sound velocity of lSOlmls,
when the vessel rolls. During the roll tests, the near- lm/s highcr than measured. The remaindcr of the water
surface sound velocity gradient was on the order of column is iso-velocity at 1500m/s.
lm/s/ni and the vertical motion of the transducer was
about MSm, thus the induced depth error to be expected The near-surfhce gradient problem discussed (above is
at 3~75' (from Fig. 5 ) is about 39.4% on alternate roll somewhat unique to twin-hull vessels like Crced.
peaks. Thus was in general agreement with the Moreover, it likely would not appear under n o d survey
hydrographic data logged at the time,which showed roll- conditions for the following reasons: a) the Creed operates
synchronous errors of the same order of magnitude; with minimal roll becaw of its level-seeking, dynamic
control system; and b) in conditions rough enough to the
In our evaluation of motion sensorswe initially attributed make the vessel roll si@cantly, the surface layer would
the Creed's depth errors to inaccuracies in measuring the be well mixed by wave action, thus making the near
roll angle: an error of 0.06" in roll can also produce a surface sound velocity gradient zero, or close to it.
0.4% error in the depth measured by the 75" beam
(Loncarevic and Scherzinger). The idea of the gradient
causing the error (Hutchins) was subsequently
investigated and in light of that analysis, and a season of
field use with the POS-MV sensor, we believe the error is
properly attributed to the near-surface gradient.

1.o

0.5
3?
i
4r 0.0
-0.5

-1 .o
-5 0 5
SSV Error, 4 s

Fig. 5. Sensitivityof depth errors to surface sound velocity LTRANSDUCER


(SSV) errors for various multi-beam systems at zcro roll
angle. The surface layer (0.3m) has an actual sound
vclocity of 1501mls, lmls higher than measured. The Fig. 7. Bow-on view of survey vessel Frederick G. Crced
remainder of the water column is iso-velocity at 1500ds. showing the location of the sonar transducer
1008

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
Fig. 6 shows, that for transducers with non-zero mounting The oceanographic conditions (salinity, temperature,
(or sector) angles (EMlOOO, BCC, Sea Bat), there is very and pressure) which determine the sound velocity
little modulation of the depth error by roll under profile have temporal and spatial variability. The
conditions of I d s error in the surface sound velocity problem of achieving the required depth accuracy is
(SSV). This is so because of the limited amount of phase aggravated in near-shore areas where the influx of
controlled beam steering required with these fresh water and tidal mixing could result in rapid
transducers, even for extreme beam angles. The effect changes in SVPs over short time periods and
increases more or less in proportion with the SSV error. distances. It is important to use specific, local
In contrast, the EM3000 performance shows an oceanographic knowledge to full advantage in order to
increasing sensitivity to roll at beam angles of 75". choose the minimum distance between SVP stations
Note that 75" is outside the normal operating band for and the maximum period of time between repeat
EM3000; calculation shows, however, for a +70" ray profiles.
angle, -10" roll and I d s SSV error, that the depth
error is about 0.5%. A close examination of the above constraints may well
indicate that in difficult areas, stopping the ship to
characterize the sound velocity in space and in time
4. CONCLUSIONS will leave too little time for actually surveying. In
such a case, one solution is to use a near-real-time,
In this paper we have shown: i) that the errors in sound velocity profiler that is deployed continuously
estimating the actual sound velocity in the vicinity of while the ship is running its survey lines. CHS is
the transducer face can lead to sigmficant errors in the currently examining the feasibility of this approach for
sea depth calculations, and ii) that the sensitivity to use on its larger vessels and on 10m survey launches.
sound velocity errors is very dependent on the amount A second approach is to use adaptive modeling of the
of phase steering needed for beam pointing control. error regime coupled with deliberately-introduced
Phase steering requirements, in turn, were shown to be redundancy in the depth data in an effort to enable
a function of the geometry and mounting angle of the interpolation between temporally and spatially sparse
beamforming elements of the transducer. Inasmuch as SVPS.
vessel roll has the effect of changing the effective
mounting angle, it too, contributes an error to the sea
depth; for flat transducers t h s error becomes highly 5. ACKNOWLEDGEMENTS
nonlinear at large beam angles.
The authors acknowledge the gracious help of E.
Vaughan in typing the manuscript, and of G. Steeves
Although particular, commercial, multi-beam sonars
and G. Henderson in reviewing the final document.
were used in our analysis, the results are applicable to
systems using three generic classes of transducer: a)
flat-level, b) flat-tilted, and c) circular-faced, sector-
6. REFERENCES
steered.

From our analysis, it follows that the survey strategy Anon, 1987. IHO Standards for Hydrographic
pursued must consider the oceanographic conditions Surveys; International Hydrographic Organization
witlvn the project area and must be adjusted to fully Special Publication No. 44, Monaco (3rd edition).
account for the transducer configuration. In addition,
due to the high sensitivity of sound velocity to Burdic, W.S., Underwater Acoustic System
temperature and salinity, it also follows that Analysis, Prentice-Hall Signal Processing Series,
continuous sound velocity measurements should be Prentice-Hall, Inc., Eagle wood Cliffs, NJ, 07632,
made at the face of the transducer; particularly in 1984.
areas of poor mixing when solar heating is strong.
Cervenka, P., U.C. Herzfield, and C. de Moustier,
With the contemporary trend in multi-beam mapping Accuracy of the Spatial Representation of the
toward the use of small, economical boats that are Seafloor with Bathymetric Sidescan Sonars; Marine
given to high pitch and roll, attention should be paid Geophysical Researches, 16: 407-425, 1994.
to the modulation of refraction errors by roll angle.
This is more of a concern with flat-level transducers at Dinn, D.F. and B.D. Loncarevic, 1994, An Evaluation
large beam angles. of Ship Motion Sensors; Proceedings of the
Kinematics Systems Conference, Aug. 30 - Sept. 2,

1009

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.
1994, Dept of Geomatics, University of Calgary, 7. BIOGRAPHICAL INFORMATION
Banff, Alberta, Canada.
Donald F. Dinn is a professional engineer specializing in
Dodds, D.J., Methods for Correction of Refraction electronic systems. As Manager of Engineering and
Errors in Multi Beam Echo Sounders; GeoAcoustics Technical Services at Bedford Institute of Oceanography ,
Inc., Aurora, Ontario, Canada; DFO Contract Report, he has been project engineer for various activities in
FP963-4-3015/01HAL, 1994 support of the Canadian Hydrographic Service (CHS),
including retrofitthg CHS' DOLPHIN vehicle with an
Hughes Clarke, J.E. and Godin, A., Investigation of EM-100 multi-beam sonar, the DOLPHIN handling
the Roll and Heave Errors Present in Frederick G. system, and the MATTHEW Motion Sensing Trials
-
Creed EM-1000 Data When Using a TSS-335B (MMST-93). Currently he is part of the team preparing
Motion Sensor; Ocean Mapping Group, University of to commission the Simrad EM3000 multi-beam sonar for
New Brunswick, DFO Contract FP707-3-5731, 1993. use on 10m survey launches.
Hutchins, R., Applied Analytics Corporation, Malton, Bosko D. Loncarevic a marine geophysicist interested in
Ontario, Canada. Personal Communication, 1994. measurements of Potential Fields over the oceans. His
work in dynarmc gavimetry has necessitated developing
Lachapelle, G, Lu, G. and Loncarevic, B. D., Precise an understanding of ship behaviour which has led to his
Shipborne Attitude Determination Using Wide more recent involvement with precision multi-beam
Antenna Spacing; Proceedings of the Kinematics sounders. He has been a key player in evaluating shp
Systems Conference, Aug. 30 - Sept. 2, 1994, Dept of motion sensors and in examining multi-beam sonar error
Geomatics, University of Calgary, Banff, Alberta, statistics.
Canada.
Gerard Costello has been a Hydrographic Surveyor
Loncarevic, B.D. and Scherzinger, B.M., with the Canadian Hydrographic Service since 1979.
Compensation of Ship Attitude for Multi-beam He graduated from the University of New Brunswick
Sonar Surveys; Sea Technology, 35,. (6), pp. 10-16, with a degree in Survey Engineering and has worked
June,1994. in both the Atlantic and Quebec Regions of CHS,
where he developed his expertise in hydrographic
Loncarevic, B.D., and the Shipboard Scientific Party, surveying. During the past few years, Mr. Costello
Cruise Report: MATTHEW Motion Sensor Trials has been managmg the CHS Coastal and Ocean
(MMST-93); Geological Survey of Canada Open File Mapping Program, where he introduced such new
Report No. 93-2730, 1993. technologies as the multi-beam survey systems and
associated computer-aided data processing. He has
Luscombe, J., Enhancement of Multibeam Echo- recently been awarded the DFO Deputy Minister's
sounder Performance by Use of High Accuracy Commendation Award, in recognition of his
Motion Sensors; Paper presented at HYDR094, Ninth achievements with CHS.
Biennial International Symposium, Sept. 13 -15, 1994,
The Hydrographic Society, Aberdeen, Scotland.

1010

Authorized licensed use limited to: Hong Kong Polytechnic University. Downloaded on March 08,2024 at 12:08:38 UTC from IEEE Xplore. Restrictions apply.

You might also like