BREASTING 20-knot winds and sheets of freezing rain, the Alliant tracked back
and forth a short distance from the shore of Seattle鈥檚 Lake Washington. On a
cable 20 metres beneath the waves, the flat-bottomed vessel towed a 6-metre-long
tubular pod packed with electronics. Several times each second, the pod
broadcast a pulse of high-frequency sound and listened for reflections from the
bed of the lake, a further 20 metres below. The equipment inside the 1-tonne pod
was designed to seek out a number of objects no bigger than a teapot that had
been scattered over the lake bottom. The search, last December, took three days.
Its purpose was to test a revolutionary sonar system which could yield images of
the ocean floor that are as clear as if the murky water above had
disappeared.
The technology is called synthetic aperture sonar (SAS), and it has the
potential to produce pictures of the seafloor showing detail that is 20 times as
fine as is possible with conventional sonar imagers. Experimental SAS systems
can already pick out features measuring 40 centimetres from a range of up to 400
metres, and the aim is to do much better than that. The driving force behind
development work on the new system is the US Navy, which plans to use it to hunt
for mines. But within a few years anyone who wants to know what is hiding
beneath the waves鈥攆rom oceanographers and cable companies to mineral
developers and treasure hunters鈥攃ould be working more efficiently and
effectively with descendants of the equipment in Alliant鈥檚 one-tonne pod.
On the pulse
Advertisement
All sonar systems work by sending out pulses of sound, and then listening for
the echo as they reflect back from objects in the water. Sound travels through
water at a more or less constant speed鈥攁round 1500 metres per
second鈥攕o it is possible to work out how far away the object is by
measuring the time it takes for the echo to return.
The simplest sonar setups pick up the echoes on a single transmitter and
microphone. But most modern sonar systems use a bank of underwater
transmitters-receivers to monitor the arrival time of the returning signals,
their strength and their frequency. The frequency is important when the target
is moving relative to the receiver, because the Doppler effect then comes into
play: the frequency of the echo is shifted upwards when the target is moving
towards the receiver, and downwards when it is moving away.
The big problem with conventional sonar is its poor resolution. For a
teapot-sized article lying on the seabed to be detected, the sonar system must
be 10 metres away or less; from any farther off it is unable to distinguish the
object from its surroundings. At a range of a few hundred metres, even the best
systems fail to pick out objects that are less than a few tens of metres in
size. 鈥淔rom that distance, you might be able to resolve the hull of a ship,鈥�
says David Cohen, a physicist with Dynamics Technologies, one of two companies
funded by the US government to develop SAS. 鈥淵ou certainly couldn鈥檛 resolve
objects the size of a teapot.鈥�
The term 鈥渁perture鈥� in the name refers to the length of the array of
transmitter-receivers which stretch for several metres underwater. The longer
this array, which may contain as many as three dozen units, the better the
resolution. The size of the aperture puts a fundamental limit on the size of the
objects that a conventional sonar system can pick out, in the same way that the
resolution of an optical telescope depends on its diameter.
In theory, improving the resolution is easy: simply lengthen the array. But
in practice, longer arrays are complex, expensive and unwieldy. Instead,
engineers have adapted a technique that is already used in radar: they are
trying to create the illusion of a long aperture by moving the array through the
water and then combining the data from several snapshots taken from different
positions. The distance the array moves during this process is known as the
synthetic aperture, and in theory there is no limit to the resolution it can
achieve. 鈥淲ith a synthetic aperture system, you could create an aperture 5
metres long or 100, just by moving the array farther,鈥� says Kenneth Rolt, an
engineer with Sanders, a defence contractor based in Nashua, New Hampshire.
But while the principle is simple, the practical difficulties are huge. To
produce an image, the computer must combine data, such as the range and Doppler
shift of the return signal, from a succession of snapshots. To understand how it
does this, imagine the footprint of a sonar system on the seafloor and that a
point in this area is occupied by a target. As the sonar moves past, both the
Doppler shift of the return signal and the distance of the target change.
Predicting how is a relatively simple mathematical task for a computer and
allows it to combine the signals from successive snapshots.
Pixelated image
The process is complicated by the fact that the computer does not know where
the target is, and so must repeat this procedure for every pixel in the image.
According to Cohen, for a relatively small area of seafloor, say 500 metres by
500 metres, and a resolution of 10 centimetres, this sequence of steps must be
applied 25 million times. The same approach has been successfully used for radar
imaging for more than a decade.
But transferring the technique from radar to sound imaging is far from
straightforward. While radar pulses are not greatly affected by variations in
air temperature and density, the temperature and salinity of seawater can
interfere with acoustic signals. And a sonar system cannot simply be bolted to
the belly of a ship in the way that a radar set can be mounted on an aircraft. A
ship鈥檚 metal hull can ring like a bell when tapped by an acoustic pulse, and the
slap of the waves also creates a din that interferes with the echoes.
So far, the most effective way to isolate a sonar array from the noise on the
surface is to place it in a pod and tow it behind a ship, below the surface of
the water. But this creates problems of its own. The array of transponders is
constantly battered by currents, and each toss of the boat whips the cable that
links them. 鈥淐ompensating for that motion so the system thinks it鈥檚 travelling
in a straight line is the biggest problem remaining in SAS,鈥� says James
Christoff, a physicist with the US Navy鈥檚 Surface Warfare Center鈥檚 Coastal
Systems Station in Panama City, Florida, and a leader in high-resolution SAS
research.
Cohen agrees. 鈥淚n order to process sonar information accurately, we need to
know the distance between the towpod and the object that鈥檚 sending back echoes.
To do that, we need to know the exact position of each of a system鈥檚 receivers
at any given instant鈥攁nd we need to know it to within a fifth of a
wavelength of the acoustic pulse you鈥檙e transmitting.鈥� If the returning echo is
miscalculated by even half a wavelength, the data become meaningless. 鈥淕iven the
frequencies we use, that means we have to be able to plot each receiver鈥檚
position to within a few centimetres of the theoretical straight line it鈥檚
travelling along,鈥� Cohen explains. This is no easy task when the array can
zigzag several metres either side of this notional straight line.
Conventional sonar systems, as well as most SAS systems under development,
use inertial motion detectors such as accelerometers to measure changes in
attitude and position of the transponders. The data they generate can then be
used by the system鈥檚 computers to plot any deviations from the ideal straight
line. This can then be taken into account when calculating the image from the
returning pulses. 鈥淚n processing the data, software can correct the readings and
make them look like the towpod actually did travel in a straight line,鈥� explains
Cohen.
Though this concept, too, is borrowed from well-tested methods used for
synthetic aperture radar, the problems are magnified when it comes to SAS.
Accelerometers cannot measure changes in motion perfectly. But Cohen says this
hardly matters when a fast-flying aircraft is sending out pulses that travel at
the speed of light, because collecting and integrating the data takes only a
second or so. Acoustic pulses and the ships producing them move relatively
slowly, however, and collecting the data can take several minutes. During that
time errors build up. 鈥淭hose biases can completely skew your readings,鈥� says
Cohen. Engineers at Dynamics Technologies have calculated that an accelerometer
for high-resolution SAS must be 200 times as sensitive to changes in
acceleration as one used for synthetic aperture radar. Researchers are having a
tough time trying to build accelerometers that can achieve this sort of
accuracy.
Others think the quest for such devices is futile. 鈥淗ardware only gets you so
far,鈥� cautions Peter Gough, an electrical engineer at the University of
Canterbury in New Zealand, who has been working on SAS since 1983. 鈥淭o get
resolution down to the level of centimetres, eventually you have to correct the
data itself.鈥�
To do this, Gough and other scientists have developed a mathematical
technique called autofocusing. The idea is simple, he says: 鈥淭he seafloor is a
random distribution of bits and pieces, but movement in the towpod creates an
error across the entire image.鈥� If the pod moves toward the echoes, the echoes
appear to arrive sooner then they should, making that part of the seafloor seem
closer. If the pod retreats from the echoes, the seafloor seems further
away.
Autofocusing works by looking for this type of error. The computer first
identifies any features that run across the length and breadth of the image,
then simply shifts the echoes backward or forward in time to see what happens.
鈥淲hen an optometrist examines your eyes, he tries one lens at a time to see
whether it makes your vision better or worse,鈥� says Rolt. 鈥淚n sonar processing,
the software moves the returned echo a little forward or a little backward in
time to see which sharpens the image. It keeps moving the data in the direction
that clarifies the picture until the image stops getting sharper.鈥�
Promising results with autofocusing are raising hopes that it might also help
to correct readings skewed by variations in the water itself. 鈥淭here are cool
spots and warm spots, and sometimes both where currents join,鈥� Christoff
complains. 鈥淭he salinity changes, especially where freshwater flows into
brackish. All of these factors affect the speed of sound through
water鈥攚hich, in turn, changes the appearance of the data you collect.鈥�
Researchers expected this to be a big problem. 鈥淭here have only been six
experiments done to study the effects of ocean water鈥檚 instability on sonar,鈥�
notes Gough, 鈥渁nd our lab did one of them.鈥� Gough and his team transmitted
acoustic pulses through water that varied vastly in temperature and salinity,
and found that all the pulses seemed to travel at almost identical speeds. 鈥淎t
first, I didn鈥檛 believe our results,鈥� says Gough. 鈥淭hey were just too good.鈥� He
was finally convinced when he saw similar results from another group at a
conference in 1993. Christoff explains: 鈥淥ur equipment isn鈥檛 able to distinguish
between errors created by motion and errors created by the instability of
water,鈥� he says.
So just how good is SAS? Some experimental versions of SAS can resolve a
10-centimetre object from 400 metres. Another system is designed to map the
seabed with centimetre accuracy from 50 metres, while an even more ambitious
plan aims to pinpoint 5-centimetre targets across distances of a kilometre or
so. 鈥淎nything you can do with conventional sonar can be done more effectively
with SAS,鈥� Cohen predicts.
Charting the depths
Phone companies laying telecoms cables or petroleum companies laying undersea
pipes to carry oil and gas use sonar to chart the smoothest path for their
lines. Exploration companies need to find a flat area of seabed on which to
place their oil rigs before they can drill in search of oil. And scientists want
to make detailed maps of coastal seabeds.
While these groups get by without the ability to recognise a teapot at 1000
metres, all would like to cut the time they need to keep a sonar-rigged ship at
sea鈥攁t a typical cost of up to $20 000 a day. One advantage of SAS
systems is that their field of view is as much as 20 times that of conventional
systems. This could lead to a similar reduction in the time and cost of being at
sea.
There are times, however, when teapot-size details are vital. Cohen cites a
recent attempt to clean up a radioactive waste dump in the sea off
Massachusetts. 鈥淩adioactive waste had been dumped over many
years鈥攕ometimes in large drums, but sometimes in buckets,鈥� he explains.
鈥淚n a situation like that, you not only want to cover as wide an area as fast as
possible, you also need to be able to resolve fine detail. That鈥檚 the kind of
application that SAS is designed for.鈥�
According to Gough, the biggest single barrier remaining to the rapid
commercialisation of SAS is human rather than technological. 鈥淲ithin a year or
so, all the technical problems will have been dealt with,鈥� he forecasts. 鈥淰ery
soon, we鈥檒l be able to produce images of the seafloor that have the same impact
as if you鈥檇 been able to strip away the water and take an aerial photograph.
What we need to do now is to learn to understand what we鈥檙e beginning to see for
the first time.鈥�
