Scientists: Acoustic episodes could be bizarre accident


“Acoustic interference without malicious intent” could have caused brain injuries to American diplomats in Havana, according to a study published earlier this year.
Under one possible scenario, a covert ultrasonic eavesdropping device could have interfered with an ultrasonic jamming device, creating the “strange metallic sounds” that caused injury.
Ultrasonic signals are not audible, but can be used to capture audio data. A computer is then used to process the silent ultrasonic signal, converting it to audio that a person can hear.
Researchers from the University of Michigan and Zhejiang University in China studied what happens when ultrasonic signals clash, creating what’s known as intermodulation distortion, or IMD. They published their findings in the journal, “Computers in Biology and Medicine.”
Below are excerpts of their study, “On Cuba, diplomats, ultrasound, and intermodulation distortion.”
This paper analyzes how ultrasounds could have unintentionally led to the AP news recordings of metallic sounds heard by diplomats in Cuba. Beginning with screen shots of the acoustic spectral plots from the AP news, we reverse engineered ultrasonic signals that could lead to those outcomes as a result of intermodulation distortion with non-linearity in the acoustic transmission medium. We created a proof of concept ultrasonic device that amplitude modulates a signal over an inaudible ultrasonic carrier. When a second inaudible ultrasonic source interfered with the primary source, intermodulation distortion created audible byproducts that share spectral characteristics with audio from the AP news. Our conclusion is that if ultrasound played a role in harming diplomats in Cuba, then a plausible cause is intermodulation distortion between ultrasonic signals that unintentionally synthesize audible tones. In other words, acoustic interference without malicious intent to cause harm could have led to the audible sensations in Cuba.

In early 2017, diplomats in Cuba suffered hearing loss and brain damage after hearing strange metallic sounds. The news media published reports ranging from scientific analysis of sound recordings to the diplomatic implications. The mystery deepened after physicians published two dueling JAMA papers on neurological damage to diplomats. The news media remained flummoxed on what may have caused the neurological damage. Several news reports suggested that an ultrasonic weapon could have caused the harm. Other experts suggested toxins or viruses. The cause remains a mystery. The substantiated facts include:

  • Ultrasonic tones are inaudible to humans.
  • Diplomats in Cuba reported hearing audible sounds.

Therefore, any sounds perceived by diplomats are not likely the ultrasound itself. We were left wondering:

  1. How could ultrasound create audible sensations?
  2. Why would someone use ultrasound in the first place?

Why Ultrasound.
It is well known that audible sounds typically propagate omnidirectionally and are difficult to confine to parts of a room. In contrast, ultrasounds tend to propagate within a narrower beam than audible sound and can focus a beam towards a more specific area. News reports cited diplomats discussing sounds that were narrowly confined to a room or parts of a room. This type of observation is strongly correlated with ultrasound. We believe that the high-pitched audio signals confined to a room or parts of a room are likely created by ultrasonic intermodulation distortion.

How to Produce Audible Sound from Ultrasound.
Humans cannot hear airborne sounds at frequencies higher than 20 kHz, i.e., ultrasound. Yet the AP news reported that “It sounds sort of like a mass of crickets. A high-pitched whine, but from what? It seems to undulate, even writhe.”
The AP’s spectrum plot shows a strong audible frequency at 7 kHz. We believe that this 7 kHz sound is caused by intermodulation distortion, which can down-convert the frequency of ultrasound into the audible range—resulting in high-pitched noises. Nonlinearity typically causes intermodulation distortion. The engineering question boils down to: assuming an ultrasonic source, how can the audible byproducts consist of a mixture of several tones around 7 kHz separated by 180 Hz, as described by the AP news recording?

Sources of Ultrasound.
There are many potential sources of ultrasound in office, home, and hotel environments. Energy efficient buildings often use ultrasonic room occupancy sensors in every room. Ultrasonic emitters can repel rodents and other pests with powerful ultrasonic and near-ultrasonic noises. HVAC systems and other utilities with pumps or compressors can vibrate entire buildings. Certain burglar alarm sensors, security cameras, and automated doors use ultrasound for detection of movement. Researchers from Illinois recently proposed using specially crafted ultrasound to jam microphones, and there have already been commercial ultrasonic jammers that prevent audio recording by emitting strong ultrasound noises to interfere with the microphones.
Ultrasounds are also used in offices for inaudible and location-restricted communications and control between different devices. An example is the Cisco Proximity system that utilizes ultrasound to pair a personal device with the video endpoint in a conference room so that meeting content can be shared on that device. The inaudibility of ultrasound may be used in other emerging applications that are required to be wireless and quiet, and may also be adapted for stealthy scenarios unknown to the general public, such as eavesdropping.
There are also hailing devices such as the Long Range Acoustic Device (LRAD) that many people claim use ultrasound. There may be LRADs that use ultrasound, but modern LRADs tend to use parametric audible sound below 3 kHz. Using an array of several dozen piezo speakers that emit sound in a synchronized fashion to improve directionality, a LRAD can generate sound waves with a wavelength much smaller than the size of the speaker. Under such conditions (which also tend to be true of ultrasonic emissions), the sound will propagate in a tight, directional beam—enabling long distance delivery of sound.
If sounds from an ultrasonic source were to collide with another such source, the two signals could combine to form audible byproducts in both air and microphones due to nonlinearity. The interference between ultrasonic devices is common because many such devices adopt the same frequency band of ultrasound, generally between 20 kHz and 40 kHz. Compared with RF signals, the bandwidth of ultrasounds is narrow, and there is no spectrum allocation schemes or mechanisms that detect collision on existing ultrasonic devices.

Assumptions and Limitations.
We assume that the sound came from ultrasound, then work backwards to determine the minimal characteristics of an ultrasonic source that would explain the observed audible sensations.
We assume the recordings from AP news are authentic, but remain skeptical because we are unaware of where and how these recordings were made. There could be added distortion in the AP audio, so we cannot assume the recordings reflect what humans actually perceived. In one video, the AP news is seen playing a sound file from one iPhone to a second iPhone, essentially making a recording of a recording. Each traversal through a speaker or microphone will add distortion and filtering.

Spectral analysis of AP news audio
We initiate our study with two observations from the AP news: (1) the original audio recordings and (2) description on the high-pitched sounds heard by those in Cuba. Our goal is to construct ultrasonic signals that can lead to similar spectral and audible characteristics.

Audio Clips.
The AP News published several recordings from Cuba described as a high-pitched whine or “cricket” sound. In the video, a piercing, metallic sound is evident which is not pleasant to hear. As a common method to analyze signals, a frequency spectrum was obtained by Fourier transform of the original sounds. The AP news performed the spectral analysis on a smartphone and showed a spectral plot centered at 7 kHz. The spectral plot demonstrates that there are roughly 20 or more different frequencies embedded in the audio recording. Watching the AP video frame by frame, we immediately noticed a few oddities. In one sequence, someone plays a sound file from one smartphone while a second smartphone records and plots the acoustic spectrum. Thus the data may be significantly corrupted, because each microphone and speaker introduces some distortion. Moreover, what humans hear isn’t necessarily the same as what a microphone detects. Cleverly crafted sounds can lead to auditory illusions to microphones akin to optical illusions.
Nevertheless, we decide to begin our analysis from the AP news audio clips. We considered the question: given its authenticity, what source can produce the sound while satisfying the description provided by the personnel in Cuba? Firstly, we compare the recording to that of a similar sound, i.e., that of cicada vocalization. The AP news recording is similar to the call of cicadas, because of the overlap at the 7 kHz frequency band, and the presence of multiple pitches with separation. Despite the similarity, the sound patterns have substantial differences as well. For instance, swarms of cicadas are not known for their ability to perform collective acoustic beam forming or phased arrays to create localized sound.
Neurologists who examined the injured diplomats published their findings in JAMA, and suggest that the neurological damage is real. However, there are limitations to the retrospective study, namely, causality is difficult to establish without a control trial or elimination of other null hypotheses. Our report does not itself contribute any new findings on neurological harm.

Alternative explanations
While our results do not rule out other potential causes, the results suggest that ultrasound without harmful intent could have led to accidental harm to diplomats in Cuba.
We originally suspected subharmonics of ultrasound as the cause, but this hypothesis would not align well with the spectral analysis provided by the AP news. Rather than evenly spaced ripples in the frequency domain, we would expect to see frequencies at 1/n submultiples of the fundamental frequency for integers n, if subharmonics were responsible for the observed patterns.
180 Hz is at the high end of the fundamental frequencies of average male conversational voices. It may be coincidence that the tones are 180 Hz apart, but this could also indicate a type of voice eavesdropping modulated over ultrasound and gone awry.
While the mathematics leads us to believe that intermodulation distortion is a likely culprit in the Cuban case, we haven’t ruled out other null hypotheses that may account for the discomfort that diplomats felt. For example, perhaps the tones heard by personnel didn’t cause their symptoms but were yet another symptom, which may be indicative of the actual cause. Or perhaps the sounds had a non-auditory effect on hearing and physiology, via bone conduction or some other phenomenon. Microwave radiation is another possible source of the health effects. Pulsed radio frequency energy can cause an auditory response within the human head due to the thermoelastic expansion of portions of the auditory apparatus. However, a remaining question is whether microwaves could have produced the highpitched sounds recorded by the smartphone in the AP news video.

Related work
The notion of using audible and inaudible sound to cause auditory and sensory illusions is not new. Our results build upon the following research.
In our past research, we use audible and ultrasonic tones to test the cybersecurity of computer systems.
We have urged more attention to be paid to the physics of cybersecurity, and the events in Cuba provide more evidence of the need
to understand the causal relationships between physics and cybersecurity.

Unresolved questions
Our report shows how simple ultrasonic tones and intermodulation distortion can provide an explanation for the reported symptoms of hearing loss, localized sound, spectral patterns, and the ability for a diplomat to record the sounds. Our approach differs from other hypotheses, in that we implemented hypothesis-testing experiments, and we supposed accidental harm rather than intentional harm when developing our test system. This allowed us to consider the effects of highpower ultrasonic transmissions from within a diplomat’s home rather than the more difficult question of how to beam ultrasound into a home from a great distance. However, our experiments do not eliminate other hypotheses. In particular, several mysteries remain:

  • How could ultrasound penetrate walls into homes and offices? Could an emitter be outside the premises or planted inside? Was it primarily airborne, or did it originate as a contact vibration?
  • At what level of intensity could IMD products cause harm to humans? We know of no non-trivial lower bounds. Based on our
    reading of various safety documents, we believe most countries set conservative thresholds for airborne ultrasound out of an abundance of caution and to compensate for uncertainty. While there are anecdotes and folklore concerning harm from airborne ultrasound, we have found no primary sources that confirm this, aside from stories about extremely intense sounds above 155 dB.
  • What about standoff distance? Our report does not investigate distance beyond 5 m. We do not have a facility to safely test high intensity ultrasound, but might consider it in the future if facilities are available to do so.
  • Could audible tones be a symptom or cause? Without a controlled study, it would be difficult to distinguish a cause from a symptom. It is possible that the audible sensations are byproducts from contact vibration or some other ultrasonic source.

Study authors Chen Yan, Kevin Fu and Wenyuan Xu concluded:
Two inaudible ultrasonic tones from one or more mixing signals traveling in a nonlinear medium could easily lead to an audible intermodulation distortion product. Although little is known about how audible sound waves can cause neurological damage, rather than merely be correlated with neurological damage, the safety community has studied the mechanism by which certain audible sounds can cause pain and hearing damage. Ultrasonic intermodulation distortion can produce harmful, audible byproducts. The safety warnings on audible frequencies and intensities would apply to these byproducts.
While our experiments do not eliminate the possibility of malicious intent to harm diplomats, our experiments do suggest that whoever caused the sensations may have had no intent for harm. The emitter source remains an open question, but could range from covert ultrasonic exfiltration of modulated data to ultrasonic jammers, or even perhaps the presence of ultrasonic pest repellents. It is also possible that someone was trying to covertly deliver data into a localized space using ultrasound to say, activate a sensor or other hidden device. Our experiments suggest that tones modulated on an ultrasonic carrier by one or more parties could have collided invisibly to produce audible byproducts. These audible byproducts can exist at frequencies known to cause annoyance and pain. Another possibility would be that solid vibration (e.g., unwittingly standing on a covert transmitter) at ultrasonic frequencies for prolonged periods—leading to bodily harm. In such a case, audible intermodulation distortion could represent a harmless side effect rather than the cause of harm. Although our tests focus on frequencies rather than amplitudes or distances, we believe that high amplitude ultrasonic signals could easily produce high amplitude audible signals as unintentional byproducts, capable of harm to hearing.

Leave a Comment