Programming note: the subscribe link was broken for a while because I am
bad at computers (yet another case of “forgot to enable the systemd unit”).
It’s fixed now. The unsubscribe link was also broken and is now fixed but,
you know, maybe that was a feature. Did wonders for reader retention.
You may have seen some recent press coverage about events surrounding the
Titanic and another notable loss at sea. I’m not going to rehash much of
anything around the Titan because it’s sort of an exhaustively covered topic
in the mainstream press… although I will defend the Logitech controller by
noting that Playstation-style controllers are extremely popular interfaces in
robotics and 3D navigation (two symmetric analog sticks, unlike other major
game controllers), and considering the genuine PS4 controller’s terrible
Bluetooth pairing UX with non-Playstation devices, the Logitech is probably a
more reliable choice. And they did have spares on board!
I actually want to talk a bit about remote sensing, but of a rather different
kind than I usually mention: hydrophones and wide-area sonar. This
little-discussed military surveillance technology played a major role in the
saga of the Titan, and it’s one that seems poorly understood by both
journalists and internet randos. I’ve seen a lot of Bad Takes about the Navy’s
involvement in Titan and I want to suggest a few things that might cause you
to interpret the situation differently.
Submarines are very difficult to detect. This is a bad property for tourist
ventures to the deep sea, but a very useful property to the military. Further,
radio communications underwater are extremely difficult. Salt water attenuates
radio signals very quickly, and while the effect decreases as you go to lower
frequencies, it never goes away. Even the US Navy’s sophisticated VLF systems
require submarines to be relatively close to the surface (or rather use a wire
antenna relatively close to the surface) for reception—VLF signals only
penetrate seawater by up to about 40 meters. ELF offers better penetration into
hundreds of meters, but ELF facilities are extremely expensive to build and
operate and the receive antennas are formidably large, so the US Navy retired
its ELF infrastructure in 2004.
For this reason, submersibles like Titan communicate with their surface
support vessels via acoustic modems. This method is surprisingly reliable but
produces a very low bitrate, thus the limitation of text messaging. Similar
technology is used in deep-sea oil exploration, Titan likely used a
commercial product for the data link.
The thing that propagates best underwater, in fact far better than above water
and even better as you get deeper, is sound. The potential of sound for
detecting and locating submarines is well-known. The first prominent use of
this approach, widely called sonar, came about during the First World War when
an anti-submarine surface ship successfully detected a submarine directly below
it via reflected sound. This type of sonar works well for locating nearby
submarines, but it is an active technique. That is, an active sonar must
emit a sound in order to receive the reflection. This is actually quite
undesirable for many military applications, because emitting a sound reveals
the presence (and with sufficient receiving equipment, location) of the sonar
device. Anti-submarine ships stopped using active sonar on a regular basis
fairly quickly, since it prominently advertised their presence to all of the
submarines in the area.
Much more appealing is passive sonar, which works by listening for the sounds
naturally created by underwater vehicles. With a sensitive directional
hydrophone (an underwater microphone), you can hear the noise created by the
screws of a submarine. By rotating the directional hydrophone, you can find the
point of peak amplitude and thus the bearing to the submarine. This basic
submarine hunting technique remains the state of the art today, but the receiving
equipment has become far more capable and automated.
There is an arms race here, an arms race of quietness. I am resisting here the
urge to quote the entire monologue from the beginning of The Hunt for Red
October, but rest assured that [the Americans] will tremble again, at the
sound of [the Soviet’s] silence. In practice the magnetohydrodynamic propulsion
technology depicted on the Red October has never proven very practical for
submarines, although it was demonstrated in one very futuristic surface vessel
built by Mitsubishi and called Yamato 1 (fortunately it fared better than the
battleship by that name). Instead, the battle of submarine silence has mostly
revolved around obscure technical problems of fluid dynamics, since one of the
loudest noises made by submarines is the cavitation around the screw. I don’t
know if this is true today, but at least years ago the low-noise design of the
screw on modern US submarines was classified, and so the screw was covered by a
sheath whenever a submarine was out of the water.
Passive sonar can be performed from ships and even aircraft-deployed buoys, but
for the purpose of long-term maritime sovereignty it makes sense to install
permanent hydrophones that function as a defensive perimeter. Just such a
system was designed in the 1950s by (who else?) AT&T. AT&T had the expertise
not only in acoustic electronics, but also undersea cable laying, a key
component of any practical underwater surveillance system. Large arrays of
hydrophones, spaced along cables, were laid on the ocean floor. The sounds
detected by these hydrophones were printed on waterfall diagrams and inspected
by intelligence analysts, who relied on experience and no small amount of
educated guessing to recognize different types of marine life, geological
phenomena, and vessels at sea.
This system, called SOSUS for Sound Surveillance System, remained secret until
1991. The secrecy of SOSUS is no great surprise, as it was one of the most
important military intelligence systems of the Cold War. It presented a problem
as well, though, as few in the Navy were aware of the details of the system and
ship crews sometimes felt the abbreviated, zero-detail intelligence messages
from SOSUS to be confusing and unreliable. They were being told of likely
submarine detections, but knowing nothing about the system they had come from,
they didn’t know whether or not to take them seriously.
By the 1960s, SOSUS consisted of hundreds of individual hydrophones installed
in long, cable-tethered arrays. Cables connected the hydrophone arrays to
highly secured terminal facilities on the coast, which the Navy explained with
a rather uninspiring cover story about undefined survey work. Over the
following decades, computers were applied to the task, automatically detecting
and classifying acoustic signatures. This early automation work inspired
significant research and development on signal processing and pattern matching
in both the military and Bell Laboratories, creating early precedents for the
modern field of machine learning. Additionally, computer and telecommunications
advancements allowed for remote control of the arrays, significantly reducing
the staff required for the program and leading to the eventual closure of many
of the terminal naval facilities.
In 1984, SOSUS was renamed to IUSS, the Integrated Underwater Surveillance
System. This new name reflected not only the increasing automation, but also
the inclusion of several surface vessels in the system. These vessels,
initially the USNS Stalwart and USNS Worthy, functioned as mobile IUSS
arrays and could be moved around to either expand coverage or provide more
accurate locating of a suspected target.
The existence of IUSS was finally declassified in 1991, although it was well
known before that point due to several prominent press mentions. Since the
declassification of IUSS it has enjoyed a dual-use role with the scientific
research community, and IUSS is one of the primary sources of hydrophone
data for marine biology. Today, IUSS automatically detects and classifies
both submarines and whales.
The potential of passive sonar systems to detect submarine accidents is
well-known. The 1968 loss of Soviet submarine K-129 was detected by SOSUS,
and the location estimate produced by SOSUS facilitated the recovery of K-129
by the Hughes Glomar Explorer, one of the most fascinating naval intelligence
operations of American history. 1968 was a bad year for submarines with four
lost with all hands, and SOSUS data was used to locate at two of them (Soviet
K-129 and US Scorpion. French Minerve and Israeli Dakar would not be
found for decades).
This all brings us to the modern era. Titan was lost on, presumably, the
18th of June. It was not located on the sea floor until the 22nd, four days
later. Press reporting after the discovery included a Navy statement that
IUSS had detected and located the implosion.
This has lead to a somewhat common internet hot take: that the Navy had
definitive information on the fate of Titan and, for some reason, suppressed
it for four days. I believe this to be an unwarranted accusation, and the
timing of the location of the wreck and the statement on IUSS are readily
First, we must consider the nature of remote sensing. Remote sensing systems,
whether space-based or deep underwater, produce a large volume of data. The
primary source of actionable information in modern real-time remote sensing
are computer systems that use machine learning and other classification
methods to recognize important events. These computer systems must be trained
on those events, using either naturally or artificially created samples, in
order to correctly classify them. A major concern in naval intelligence is the
collection of up-to-date acoustic signatures for contemporary vessels so that
IUSS can correctly identify them.
A secondary method is retrospective analysis, in which human intelligence
analysts review historic data to look for events that were not classified by
automation when they occurred. Retrospective analysis, particularly with new
signature information, can often yield additional detections. Consider the case
I have previously discussed of the Chinese spy balloons: once signature
information (almost certainly RF emissions) were collected, retrospective
analysis yielded several earlier incidents that were not detected at the time
due to the lack of signatures.
Like the RF spectrum, the ocean contains a lot of noises. They come from
wildlife, from geological processes, and from commercial shipping, all besides
naval operations. The Navy does not rigorously investigate every sound
underwater, it can’t possibly do so.
When the Navy became aware of the missing Titan, analysts almost certainly
began a retrospective analysis of IUSS data for anything that could indicate
its fate. They apparently detected loud noises and were able to locate the
source as near the Titanic wreckage, probably fairly quickly after the
Titan was first reported missing.
Here is the first challenge, though: the Titan was a new submersible of novel
(if not necessarily well thought out) construction. The Navy has some
familiarity with the acoustic signatures of imploding military submarines based
on incidentally lost submarines and, in at least one case, the intentional
torpedoing of a submarine to record the resulting acoustics (the Sterlet).
This data is used to produce a signature against which new signals can be
compared. Because of the significant differences in size and construction
between Titan and military submarines, the Navy likely had very low
confidence that known acoustic signatures of catastrophic losses were
applicable. The total number of submarines to have ever imploded underwater is
quite small, and none were of similar size and construction to Titan. The
point is that while intelligence analysts likely suspected they had evidence
of implosion, they probably had low confidence in that conclusion.
It is unwise, in the course of a search and rescue operation, to report that
you think the vessel was irrecoverably lost. Doing so can compromise search
operations by creating political pressure to end them, while making the
situation of families and friends worse. It is customary to be very cautious
with the release of inconclusive information in events like this. The problems
are exemplified by the Coast Guard’s announcement that another passive sonar
system had detected possible banging sounds, which motivated a lot of reporting
making wild conclusions based on acoustic signatures that were likely
The more damning accusation, though, is this: did the Navy withhold information
on the detection from searchers out of concern for secrecy? Setting aside that
this makes little sense considering that SOSUS and its capabilities have been
widely known to the public for decades, and the search site was well within
historically published coverage estimates for SOSUS, this accusation doesn’t
align with the timeline of the search.
The first search vessel capable of deep undersea exploration, the ROV Pelagic
Odysseus 6k, arrived on the scene on the morning of the 22nd. Just five hours
later, Odysseus had located the wreckage. Considering that the descent to depth
alone would have taken Odysseus over an hour, the wreckage was located extremely
quickly in the challenging undersea environment. One reason is obvious:
the wreckage of Titan was close to the Titanic, although the Titanic debris
field is large and searching it all would have taken hours. The second reason
became known shortly after: when Odysseus began its search, they had almost
certainly already been tipped off by the Navy as to the location of the possible
The Navy did not withhold information on the detection for four days out of some
concern for secrecy. Instead, the information was not known to the public for
four days because that was when the search team was first able to actually
investigate the Navy’s possible detection.
Indeed, the idea that the Navy suppressed the information seems to come only
from the rumor mill and internet repetition of half-read headlines. The
original press coverage of the IUSS detection, from the WSJ, states that the
Navy reported the finding to the Navy commander on-scene at the search effort
immediately. It does include the amusing sentence that “the Navy asked that the
specific system used not be named, citing national security concerns.” This
might seem like a huge cover up to those unfamiliar with intelligence programs,
but it’s perfectly in line with both normal military concerns around classified
systems (which are often known by multiple names which must be kept
compartmentalized for unclassified contracting) and the specific history of
IUSS, which during its period of secrecy had problems with being accidentally
named in unclassified reports multiple times.
IUSS is now a smaller system than it once was, although with improving
technology its coverage has probably expanded rather than contracted. It still
serves as a principal method of detecting submarines near the US, an
important concern since submarines are one of the main delivery mechanisms
for nuclear weapons. IUSS is just one of several semi-secret underwater sensing
systems used by the Navy.
A not totally related system that will nonetheless be of interest to many of my
readers (who I suspect to be somewhat concentrated in the San Francisco Bay
Area) is the San Francisco Magnetic Silencing Range. A small building in the
parking lot of Marina Green, complete with a goofy little control tower from
the era of manned operation, is the above-water extent of this system that uses
underwater magnetometers to measure the magnetic field of Navy vessels passing
through the Golden Gate. Since underwater mines are often triggered by
magnetometers, the Navy ensures that the magnetization of vessel hulls does not
exceed a certain limit. If it does, the vessel can be degaussed at one of
several specially-equipped Navy berths—inspiration for at least one episode
of The Next Generation. Similar arrays exist at several major US ports.
The building itself is long-disused, and the array is now fully remote
controlled. When I lived in San Francisco it was abandoned, but I see that it
has apparently been restored to function as the harbormaster’s office. I
appreciate the historic preservation effort but something is lost with the
removal of the Navy’s sun-faded signage.