Electronics Guide

What a Body Area Network Is

A body area network connects a handful of sensors, actuators, and one coordinating hub across distances of centimeters to about two meters. The node count is small, typically fewer than a dozen. Data rates span an unusually wide range for such a compact network: a temperature patch may report a few bytes a minute, while a multichannel electromyography sleeve or a neural recording implant may need hundreds of kilobits or several megabits per second. Latency requirements diverge just as sharply. A step counter tolerates seconds of delay. A closed-loop system in which an implanted sensor triggers an implanted therapy tolerates milliseconds.

Four traffic classes recur across nearly every deployment. Periodic telemetry dominates by packet count and is the traffic that sets average current draw. Event-driven alarms are rare but must never be dropped, and they set the reliability requirement. Bulk transfer happens during a clinic visit or an overnight sync and sets the peak rate. Closed-loop control is the smallest class and the most demanding, because a late packet is a clinical event rather than a missing data point. Designing for the average of these four produces a network that fails at all of them; the usual approach is to give each class its own access mechanism.

The topology is almost always a star. One hub, which may be a phone, a bedside monitor, a chest-worn gateway, or an implanted device that outranks the others, holds the schedule and relays traffic beyond the body. IEEE 802.15.6 formalizes this as a one-hop star, with an optional two-hop extension in which a node relays for another node whose direct link to the hub is blocked. Mesh routing, which dominates building automation, buys almost nothing here: the network is too small for routing diversity to pay for its overhead, and the dominant impairment is body shadowing, which a single well-placed relay solves more cheaply than a routing protocol.

The Three Tiers of a Body Area Network

The literature divides body area networks by node placement into in-body, on-body, and off-body categories, and separately by link scope into intra-BAN, inter-BAN, and beyond-BAN tiers. The two taxonomies overlap but answer different questions: the first asks where the hardware sits, and the second asks how far a given packet travels. Both matter, because the propagation physics, the regulatory regime, and the security model all change at each boundary.

In-Body Links

An in-body node is an implant or an ingestible. Its link must cross several centimeters of muscle, fat, and possibly bone before reaching an external receiver. Tissue is conductive, so attenuation rises steeply with frequency, and the practical consequence is that implant telemetry congregates at the low end of the usable spectrum. The 402 to 405 MHz core of the Medical Device Radiocommunications Service exists precisely because those frequencies penetrate tissue acceptably while still permitting an antenna small enough to fit in a device the size of a coin.

On-Body Links

An on-body node sits against or very near the skin: a chest patch, a wrist device, a hearing aid, a textile electrode array. The path between two on-body nodes rarely goes through the body. Instead, energy travels around it, as a surface or creeping wave hugging the torso, or it reflects off the floor, walls, and furniture and arrives from outside. Both mechanisms are lossy and both vary with posture. A link from the left wrist to the right hip may be strong when the arms hang loose and much weaker when the arm crosses the chest, and the change happens within a single stride.

This variability is the defining feature of the on-body channel. It is not a fading process in the usual sense, because it correlates with the wearer's motion rather than with a random scattering environment, and it repeats on the cadence of gait. Protocols that assume a stationary channel over one packet exchange tend to work; protocols that assume one over a whole superframe tend to disappoint.

Off-Body Links

The off-body link carries traffic from the hub to infrastructure: a phone, a bedside monitor, a Wi-Fi access point, a cellular network. It is the only tier in which conventional radio engineering applies more or less unchanged, and it is also the tier where the body acts as a large, lossy obstacle in the middle of the link. A gateway carried in a trouser pocket and an access point on the ceiling may have the wearer's torso directly between them for half of every rotation.

The off-body tier is also where the network's security perimeter usually sits, because traffic that was confined to one person's vicinity becomes traffic on a shared network. Treating the three tiers as one undifferentiated network is a common design error.

IEEE 802.15.6 and Its Three Physical Layers

IEEE 802.15.6, published in 2012 as a standard for wireless body area networks, remains the only international standard written specifically for this application. Its most useful contribution is that it treats three very different physical layers as interchangeable options under one medium access layer, which forces an explicit comparison a designer would otherwise make informally. The IEEE 802.15 working group maintains an active task group, TG6ma, chartered to revise the 2012 document, so the standard should be read as a living rather than a closed specification.

The Narrowband Physical Layer

The narrowband layer is the conventional option and the one most implementations would choose. It spans seven frequency bands running from 402 MHz up to 2483.5 MHz, with channel bandwidths between 300 kHz and 1 MHz, and it delivers information rates between about 75.9 kilobits per second and about 971.4 kilobits per second. Among the bands it covers are the 402 to 405 MHz implant allocation, sub-gigahertz allocations near 863 to 870 MHz and 902 to 928 MHz, the 2360 to 2400 MHz medical body area network allocation, and the 2400 to 2483.5 MHz industrial, scientific, and medical band.

The layer uses differential phase-shift keying variants with an optional spreading factor and defines a mandatory rate per band, so any two compliant devices can at least establish a link. The design intent is sensible rather than novel: a narrow channel maximizes receiver sensitivity for a given power, and low rates suit the periodic telemetry that dominates real traffic.

The Ultra-Wideband Physical Layer

The ultra-wideband layer defines eleven channels, each 499.2 MHz wide, with center frequencies from 3494.4 MHz to 9984.0 MHz. Channel 1, centered at 3993.6 MHz, is mandatory for low-band implementations, and channel 6, centered at 7987.2 MHz, is mandatory for high-band implementations. Two signaling modes exist. Impulse radio ultra-wideband has a mandatory rate of 487.5 kilobits per second and scales up to 15.6 megabits per second. Frequency-modulated ultra-wideband has a mandatory rate of 250 kilobits per second and trades peak rate for a simpler, lower-power receiver.

Ultra-wideband is attractive here for reasons that have little to do with throughput. Its extremely low power spectral density lets it operate as an underlay beneath licensed services, its short pulses resolve multipath rather than suffering from it, and the same pulses support time-of-flight ranging, which turns a communication link into a position sensor. The cost is that these frequencies are absorbed strongly by tissue, which effectively restricts the layer to on-body and off-body use rather than implant telemetry.

The Human Body Communication Physical Layer

The human body communication layer is the most unusual part of the standard and the reason body-coupled communication belongs in this discussion. It defines a single channel centered at 21 MHz with 5.25 MHz of bandwidth. Four data rates are specified: 164 kilobits per second, 328 kilobits per second, 656 kilobits per second, and 1.3125 megabits per second, with corresponding receiver sensitivity requirements of roughly -97.35, -94.34, -91.33, and -88.32 dBm. Signaling uses a frequency-selective digital transmission scheme that spreads each symbol across the channel rather than modulating a sinusoidal carrier in the conventional sense.

The choice of a single 21 MHz channel is not arbitrary. It sits inside the frequency window, discussed at length below, where the body conducts usefully but does not radiate efficiently. It also sits below the FM broadcast band, which is the dominant ambient interferer that a body-coupled receiver must contend with. The layer's transmit electronics are strikingly simple compared with a radio: no power amplifier driving a matched antenna, no crystal-referenced carrier synthesizer of the usual kind, and no radiated emissions to certify in the conventional sense.

The Medium Access Layer and the Power-Latency Trade

The medium access layer of IEEE 802.15.6 is organized around a hub that divides time into superframes bounded by beacons. Within a superframe, the standard defines a sequence of access phases with different rules. Exclusive access phases are reserved for the highest-priority traffic, which in this context means medical emergency data. Random access phases allow contention using carrier-sense multiple access with collision avoidance in the narrowband layer, or slotted ALOHA in the ultra-wideband layer. Managed access periods carry scheduled allocations, polled allocations, and posted allocations, in which the hub grants a specific node a specific interval. A contention access phase closes out the frame.

The standard also permits operation without beacons, either with or without superframe structure, which suits networks that are almost always idle and wake only on an event. Eight user priorities map traffic onto these phases, and the mapping deliberately privileges medical and emergency data over background telemetry. This is a meaningful departure from general-purpose personal area network standards, which usually assume that all traffic is equally disposable.

The engineering substance of this structure is a trade between power and latency, and it is worth stating plainly because it recurs in every low-power network. Scheduled access is the cheapest way to move periodic data: a node knows exactly when its allocation begins, keeps its receiver off until shortly before, transmits, and sleeps again. Its average current approaches the sleep current of the silicon. The price is latency, because a node with something urgent to say must wait for its slot or contend in a random access phase. Random access inverts the trade. It responds immediately to bursty traffic but requires the node either to listen more or to accept collisions and retransmissions, both of which cost energy.

A well-designed body area network therefore uses both. Periodic telemetry rides scheduled allocations sized to the sample rate, alarms preempt through an exclusive or random access phase, and bulk transfers negotiate a long allocation when the hub is idle. Collapsing all of this into a single polling interval either wastes energy on nodes that have nothing to say or delays the one packet that mattered.

Guard intervals are where theoretical schedules meet real oscillators. A node that sleeps between allocations runs on a low-power oscillator whose frequency drifts with temperature, and the longer it sleeps, the wider the uncertainty window in which it must wake and listen. That window is pure overhead, and shrinking it means either a more accurate and more power-hungry timekeeping oscillator or shorter sleep intervals. This design calculation frequently dominates the energy budget of a low-rate node.

The Regulatory Bands That Matter

Spectrum for body area networks comes from several distinct regulatory regimes, and confusing them is a reliable way to fail certification. The following describes the United States allocations, which are the ones most often cited; other administrations have analogous but not identical arrangements.

MedRadio and the Medical Implant Communication Service

The Medical Implant Communication Service originally allocated 402 to 405 MHz for implant telemetry, divided into ten channels of 300 kHz each, with a maximum effective isotropic radiated power of 25 microwatts. That figure deserves emphasis. Twenty-five microwatts is roughly -16 dBm, which is four to five orders of magnitude below what a Wi-Fi client radiates. An implant link is therefore a link-budget exercise conducted at the edge of what a receiver can hear, and every decibel lost in tissue, in an inefficient antenna, or in a poorly chosen modulation is a decibel that must come from somewhere else.

The FCC subsequently expanded this allocation and renamed the service MedRadio. The 401 to 402 MHz and 405 to 406 MHz ranges were added in 2009. In 2011 the Commission added twenty-four megahertz spread across the 413 to 419 MHz, 426 to 432 MHz, 438 to 444 MHz, and 451 to 457 MHz ranges. The rules governing the core band constrain how a device may seize a channel, generally requiring it to monitor before transmitting so that implants do not talk over one another, with defined exceptions for emergency sessions.

Medical Body Area Networks at 2360 to 2400 MHz

In 2012 the FCC extended MedRadio to permit medical body area network devices in the 2360 to 2400 MHz band, adding forty megahertz aimed squarely at multi-sensor body networks in hospitals. The allocation is split. The 2360 to 2390 MHz segment is restricted to indoor use within health care facilities and requires coordination and registration, because it shares spectrum with aeronautical mobile telemetry that must be protected. The 2390 to 2400 MHz segment carries no such restriction and may be used anywhere, including at home.

The Wireless Medical Telemetry Service

The Wireless Medical Telemetry Service predates both and sets aside fourteen megahertz in three bands: 608 to 614 MHz, 1395 to 1400 MHz, and 1427 to 1432 MHz. It exists for cardiac and respiratory telemetry in hospitals, where interference is not an inconvenience but a patient-safety event, and it gives eligible users protected status. The 608 to 614 MHz segment carries a notable condition: users are co-primary with radio astronomy and must obtain written permission to operate within eighty kilometers of some radio astronomy facilities and within thirty-two kilometers of others. In the United States, frequency coordination for the service is administered by a designated coordinator on behalf of health care facilities, which is what keeps two hospitals in the same city from colliding.

The Unlicensed Bands

Most body-worn devices that reach consumers use unlicensed spectrum, overwhelmingly the 2400 to 2483.5 MHz band, because that is where Bluetooth Low Energy lives and Bluetooth Low Energy is what phones contain. Sub-gigahertz allocations near 868 MHz in Europe and 902 to 928 MHz in North America propagate better and are used by some proprietary body links, but they carry regional constraints, notably duty-cycle ceilings in parts of the European sub-gigahertz spectrum, that limit how often a node may transmit regardless of how much it has to report.

Body-Coupled Communication: The Body as a Lossy Conductor

The idea of using the body itself as a signal path dates to work by Thomas Zimmerman at the MIT Media Lab, published in the IBM Systems Journal in 1996 under the title "Personal Area Networks: Near-field intrabody communication." Zimmerman's system capacitively coupled picoampere currents through the body at a carrier below one megahertz, deliberately low enough that essentially no energy radiated. A demonstration exchanged electronic business cards when two people shook hands. The two properties that motivated the work then still motivate it now: the signal stays with the person, and the transmitter needs no antenna and almost no power.

Understanding why this works requires abandoning the antenna picture. At the frequencies involved, tissue is not a dielectric that a wave propagates through; it is a conductor with substantial loss. Muscle, which is mostly saline, has a conductivity on the order of a few tenths of a siemens per meter through the relevant range, and its relative permittivity is very high at low frequencies and falls steadily as frequency rises while conductivity rises. The reference measurements for this behavior are the three-part survey of the dielectric properties of biological tissues published by C. Gabriel and colleagues in Physics in Medicine and Biology in 1996, which remains the data set that simulation tools load by default.

Two consequences follow. First, an electrode pair on the skin is an electrically tiny structure heavily loaded by a lossy medium, so its radiation efficiency is negligible. Almost nothing leaves as a wave. Second, the energy that does move travels as conduction current in the tissue and displacement current in and around it, in the quasi-static regime where the phase of the field is essentially uniform over the dimensions of the body. The correct mental model is a resistive and capacitive network, not a propagating field, and the correct analysis tools are circuit models and finite-element solvers rather than ray tracing.

That model also defines the usable frequency window from both ends. At the bottom, below roughly a hundred kilohertz, the impedance of the electrode-skin interface becomes large and dominates the link, injected current runs into safety limits governing nerve and muscle stimulation, and the available bandwidth is trivially small. At the top, above roughly a hundred megahertz, the body starts to behave as a genuine antenna. It begins to radiate, which destroys the confinement that made the technique attractive in the first place, and it begins to receive, which means every broadcast transmitter within range couples into the link. The practical window therefore runs from a few hundred kilohertz to a few tens of megahertz, and IEEE 802.15.6 placed its single human body communication channel at 21 MHz squarely inside it.

Galvanic and Capacitive Coupling

Two coupling arrangements exist, and they behave differently enough that they should be treated as separate technologies rather than as variants.

Galvanic Coupling

In galvanic coupling, both transmitter electrodes contact the body directly, and the transmitter drives a differential current into the tissue. Most of that current stays in the body, spreading through the conductive volume and returning to the second electrode. The receiver, also using two contacting electrodes, measures the small differential voltage that the current gradient produces across its own electrode separation. The forward path and the return path both lie inside the body, which makes the arrangement largely independent of the surrounding environment.

Galvanic coupling favors the lower part of the window. Published guidance for compliant operation points to signals in the range of about 100 to 300 kilohertz, with injected currents on the order of one to ten milliamperes and drive voltages of one to two volts. Its characteristic weakness is distance. Because the injected current spreads through a three-dimensional conductive volume, the fraction that reaches a distant electrode pair falls steeply, and attenuation grows quickly beyond a few tens of centimeters. Its characteristic strength is robustness: nothing about the arrangement depends on how the wearer is grounded, what they are standing on, or whether they are holding a metal object.

Capacitive Coupling

In capacitive coupling, only one electrode of each device contacts the body. The other, usually called the ground or floating electrode, faces away and couples capacitively to the surrounding environment. The forward signal path runs along the body from the transmitter's signal electrode to the receiver's signal electrode. The return path runs from the receiver's floating electrode, through the air, to earth or to the surrounding conductive environment, and back to the transmitter's floating electrode. The loop closes through the room rather than through the body.

Reviews of the field consistently report that the capacitive method achieves lower path loss and higher data rates than the galvanic method, and that its advantage widens as the on-body distance grows. One comparative study of implantable configurations found capacitive electrodes delivering roughly twenty decibels more gain than galvanic electrodes at a ten-centimeter transmission distance. This is the reason nearly all higher-rate body-coupled work, including the IEEE 802.15.6 human body communication layer at 21 MHz, uses capacitive coupling.

The Return-Path Problem

The advantage comes with an awkward dependency: the capacitive link is only as good as its return path, and that return path is a capacitance to the environment measured in a few picofarads. Simple arithmetic shows why the frequency matters so much. A one-picofarad return capacitance presents a reactance of about 7.6 kilohms at 21 megahertz, but about 159 kilohms at one megahertz. The link improves by more than twenty-six decibels for that capacitance alone simply by moving the carrier up two decades. This is a large part of why 802.15.6 chose tens of megahertz rather than the hundreds of kilohertz of Zimmerman's original demonstration.

It also means the channel depends on things no data sheet describes. A wearer standing barefoot on a concrete slab has a very different earth coupling than one sitting in an office chair with rubber casters, and holding a grounded appliance changes the return path again. Measured path-loss figures for capacitive links vary widely across the literature partly for this reason and partly because measurement itself perturbs the quantity being measured.

That last point deserves care, because it has misled a great deal of published work. A network analyzer is mains-grounded. Connecting it to a body-worn electrode through a coaxial cable creates a low-impedance return path that the deployed system will never enjoy, and the resulting measurement flatters the channel by tens of decibels. Baluns and isolation transformers are the usual remedy, but the interwinding capacitance of the transformer itself can dominate the result, and investigators have reported that this parasitic influences measurements drastically. Credible characterization uses battery-powered, optically isolated, or fully floating instrumentation, and reports the grounding conditions as part of the result.

Path Loss, Channel Variability, and Electrode Design

The honest summary of body-coupled path loss is that it lands in the tens of decibels, that the spread across conditions is comparable to the mean, and that any single number quoted without its measurement conditions should be treated with suspicion. Capacitive links across a torso commonly show attenuation in the range of a few tens of decibels, which is remarkably good for a channel with no antenna and no amplifier, and which is why receiver sensitivity requirements in the 802.15.6 human body communication layer sit near -97 dBm at the lowest rate: the transmitter can afford to be very quiet.

Path loss is not a monotonic function of frequency. Below the window it rises because the electrode interface dominates; inside the window it generally falls as the return-path reactance drops; above the window radiation and ambient interference take over. Neither is it a simple function of distance. A galvanic link degrades sharply with separation, while a capacitive link degrades much more slowly, since its loss is dominated by the return path rather than by the on-body distance.

The variability sources are worth enumerating, because a protocol designer must budget for all of them:

• Posture and limb position, which change both the on-body path and the geometry of the return path, on the timescale of a stride.
• Electrode contact quality, which depends on pressure, sweat, skin hydration, hair, and whether clothing intervenes.
• Grounding of the environment, including footwear, flooring, furniture, and nearby mains wiring.
• Proximity to other conductive objects, including a second person, a metal bed frame, or a wheelchair.
• Ambient interference, which at these frequencies comes from broadcast transmitters, switching power supplies, and the wearer's own electronics rather than from other body area networks.

Electrode design is the one lever a designer fully controls. The classical instrumentation electrode is silver and silver chloride with a wet gel, which gives a low, stable interface impedance and is standard for electrocardiography, but it is unsuitable for weeks of wear because the gel dries and the skin reacts. Dry electrodes of stainless steel, gold-plated metal, conductive polymer, or conductive textile avoid that at the cost of a higher and less stable interface impedance, which matters most at the low end of the window and least at 21 megahertz. Fully insulated capacitive electrodes remove the interface chemistry entirely and shift the problem into the amplifier, whose input impedance must sit far above the coupling capacitance's reactance.

Electrode area trades interface impedance against wearability, and it interacts with placement: a larger signal electrode couples better to the body, while a larger floating electrode couples better to the environment, and the two want to be on opposite faces of the device with as much separation as the enclosure allows. For galvanic links, the separation between the two contacting electrodes sets the differential voltage the receiver sees, so wider spacing directly improves the link and directly worsens the form factor.

Energy Budgets, Harvesting, and Wireless Power

Energy is the constraint that shapes every other decision in a body area network, and the arithmetic is unforgiving. Consider an implant with a primary cell holding on the order of one ampere-hour, a typical figure for a cardiac device, and a required service life approaching a decade. One ampere-hour divided by ten microamperes of average current gives one hundred thousand hours, or about eleven and a half years. That means the entire device, including sensing, processing, therapy, and telemetry, must average roughly ten microamperes. A radio that draws ten milliamperes while transmitting may therefore be on for about one part in a thousand of the device's life, and only if it is the sole consumer, which it never is.

Worn devices face a similar calculation with different constants. A coin cell of the CR2032 class holds roughly two hundred milliampere-hours, and a rechargeable wearable battery holds tens to a few hundred. The binding constraint on a coin cell is often not capacity but internal resistance: the cell's terminal voltage sags under a transmit pulse of several milliamperes, which is why nearly every coin-cell radio design places a bulk capacitor across the cell to supply the pulse and lets the cell recharge it slowly.

The right figure of merit is energy per delivered bit, not transmit power, and for short packets the dominant term is frequently not the transmission at all. Waking a crystal oscillator, settling a phase-locked loop, and calibrating an analog front end can consume more energy than sending the payload. This is why very low-rate nodes benefit disproportionately from architectures that avoid a high-quality reference, and it is one of the genuine advantages of body-coupled signaling: a 21 megahertz baseband-like link needs far less frequency-generation machinery than a 2.4 gigahertz radio.

The same argument appears in a shipping product. Abbott's Aveir dual-chamber leadless pacemaker system uses an implant-to-implant link, marketed as i2i, in which the atrial and ventricular devices exchange subthreshold electrical pulses conducted through the blood, beat by beat. Abbott states that this conducted approach uses far less battery current than inductive, radio-frequency, or Bluetooth communication; that is a vendor claim, but it is consistent with the physics, since the transmitter is essentially a current source driving a conductive medium with no carrier synthesis and no antenna to match.

Harvesting

Harvesting is attractive because it removes the surgery or the charging routine, and it is limited because the human body is a poor power source. Thermoelectric generators exploit the gradient between skin and ambient air and produce, under favorable conditions, on the order of tens of microwatts per square centimeter; the qualifier matters, because the yield collapses in a warm room where the gradient disappears. Photovoltaic cells on a wrist or a garment work well outdoors and poorly indoors, where illuminance may be two or three orders of magnitude lower. Kinetic and piezoelectric harvesters convert motion and produce useful energy only during activity, which is precisely when many sensors most want to report. Radio-frequency harvesting works at close range from a deliberate source and is really wireless power transfer under another name. Biofuel cells that oxidize glucose or lactate remain a research subject.

None of these approaches presently powers a clinically significant implant on its own. A design that treats harvested energy as a supplement rather than a supply is far more likely to survive validation.

Deliberate Wireless Power

Transcutaneous inductive power transfer is mature and deployed. Cochlear implants have used an inductive link across the skin for decades, both to power the internal stimulator and to carry the audio data stream. The engineering issues are coil alignment, efficiency, and tissue heating, and the last is why power delivery is usually more thermally constrained than communication. Mid-field and ultrasonic power transfer extend the reach to deeper implants and remain largely in research.

Antennas Against Tissue

An antenna designed on a bench and then pressed against a body is a different antenna. Tissue has a high relative permittivity and a substantial loss tangent, and placing it in the reactive near field of a radiator does three things at once: it lowers the resonant frequency, it changes the radiation resistance, and it absorbs a large fraction of the stored energy as heat. A 2.4 gigahertz chip antenna tuned in free air routinely shifts far enough when a wrist presses against it to fall outside the band, and the resulting mismatch loss compounds the absorption loss.

The correct response is to tune with the intended loading present, using a tissue-simulating phantom or a validated simulation model, and then to check the design across the range of loading it will actually see. A wrist device sees skin on one face and air on the other; a chest patch sees skin over its whole footprint; a device in a pocket sees a thigh through fabric. Broadening the match to tolerate that range typically costs a decibel or two of peak gain, which is a bargain compared with a detuning loss of ten or more.

Implant antennas are a harder case. At 403 megahertz the free-space wavelength is about 0.74 meters, so an antenna that fits inside a device the size of a large coin is electrically minute, and it is surrounded by a conductive, lossy medium and enclosed in a metal or ceramic housing. Radiation efficiencies well below one percent are normal rather than exceptional. Designers respond by treating the implant housing as part of the radiator, feeding the antenna through a ceramic window or a hermetic feedthrough, using meandered or spiral geometries to squeeze electrical length into physical volume, and accepting that the pattern will be shaped by the body rather than by the antenna.

It is worth noting that the 25 microwatt MedRadio limit is a limit on effective isotropic radiated power, that is, on what escapes. A very inefficient implant antenna does not violate the rule; it simply forces the system to spend its scarce energy budget overcoming the inefficiency. The rule sets the ceiling and the physics sets the floor, and the link budget lives in between.

On-body antennas can be designed to exploit rather than fight the body. Planar inverted-F and patch structures with a ground plane between the radiator and the skin isolate the antenna from tissue loading and radiate outward, which suits an off-body link. A link that must travel around the torso instead benefits from a structure that couples deliberately into the surface wave. One antenna rarely serves both tiers well.

Coexistence with Bluetooth Low Energy and Wi-Fi

Any body area network that uses the 2400 to 2483.5 MHz band shares the most crowded unlicensed spectrum in the world. Wi-Fi occupies twenty-megahertz or wider channels at transmit powers many orders of magnitude above a body sensor. Bluetooth Low Energy hops across forty two-megahertz channels, Zigbee and Thread occupy sixteen, and microwave ovens radiate broadband energy near the middle of the band. A hospital ward or a modern apartment building may contain dozens of independent emitters within a few meters.

Body area networks suffer disproportionately in this environment for a specific reason: body shadowing attenuates the wanted signal while leaving the interferer untouched. An access point on the ceiling reaches a wrist sensor's receiver with no obstruction, while the sensor's own hub, in the opposite trouser pocket, is behind a torso. The signal-to-interference ratio degrades from both directions at once. This is why the FCC created MBAN spectrum just below the industrial band and why clinically significant telemetry uses protected spectrum wherever it can.

The mitigations are the familiar ones, applied with more discipline than a consumer product usually needs. Adaptive frequency hopping with an actively maintained channel map lets Bluetooth Low Energy avoid the channels a nearby access point occupies, and a device that never updates its map loses much of the benefit. Bounded retransmission with an explicit deadline beats unbounded retry, because a late clinical sample is often worthless and the retries cost energy. Scheduling around the hub's own Wi-Fi activity, where the hub contains both radios, removes a self-interference source frequently larger than anything external.

Coexistence is also a regulatory deliverable rather than an engineering nicety. The FDA recognized AAMI TIR69:2017, which describes a risk-management process for radio-frequency wireless coexistence in medical devices, and ANSI C63.27-2017, which specifies test methods and key performance indicators for evaluating wireless coexistence, as consensus standards in August 2017. A manufacturer submitting a wireless medical device is expected to present coexistence evidence built on that framework, which means defining the device's essential performance, identifying the intended radio-frequency environment, and demonstrating that performance survives representative interference.

Body-coupled links face a different coexistence question. At 21 megahertz there is no Wi-Fi to avoid; the interference comes from broadcast transmitters, from switching converters in the wearer's own devices, and from the body acting as a receiving antenna for whatever is in the air. Researchers have specifically studied how other worn devices interfere with a human body communication link, since a switching regulator a few centimeters away couples into the same electrodes that carry the wanted signal. Shielding, supply filtering, and placement are the remedies, and they are board-level problems rather than protocol problems.

Specific Absorption Rate as a Design Constraint

Specific absorption rate measures the power a body absorbs per unit mass, in watts per kilogram, and it exists because absorbed radio-frequency energy becomes heat. In the United States, the FCC limit for localized exposure from portable devices is 1.6 watts per kilogram averaged over one gram of tissue. The ICNIRP guidelines and the European regime derived from them use 2 watts per kilogram averaged over ten grams for the head and trunk, with measurement methods specified in the IEC 62209 series. Whole-body averaged limits are 0.08 watts per kilogram for the general public and 0.4 watts per kilogram for occupational exposure. The difference in averaging mass matters more than the difference in the numbers, because a smaller averaging volume captures a localized hot spot that a larger one smooths away.

For a body area network, specific absorption rate is a genuine design constraint rather than a formality, for one structural reason: the usual relief does not apply. A phone is tested at a defined separation from the body, and manufacturers exploit that separation. A body sensor has no separation. Its antenna is in permanent contact with skin, at zero distance, for the entire wear period. Whatever it radiates, it radiates into tissue.

Having said that, honesty requires a qualification. For most body area network nodes, exposure limits are not the binding constraint, because the transmit powers involved are tiny. A device radiating tens of microwatts is nowhere near 1.6 watts per kilogram, and its energy budget will run out long before its exposure budget does. Specific absorption rate becomes the binding limit in three situations: the off-body tier, where a phone or gateway radiates at hundreds of milliwatts against a pocket; wireless power transfer, where the delivered power is orders of magnitude above any communication link; and any design that combines a high duty cycle with a high transmit power to compensate for a poor antenna.

For implants the more relevant constraint is often temperature rather than absorption rate. Standards for active implantable medical devices constrain how much an implant may warm the surrounding tissue, because chronic local heating damages tissue in ways that a transient exposure calculation does not capture. The heat comes from the electronics and the battery as much as from the radio, and it is a thermal-design problem: spread the dissipation across the housing, avoid concentrating it at one face, and simulate with a perfusion model rather than an adiabatic one.

Body-coupled communication does not fit the specific absorption rate framework at all at the low end of its window. Below roughly one hundred kilohertz the dominant biological effect is not heating but stimulation of nerve and muscle tissue, and the applicable exposure guidance is expressed as an internal electric field in volts per meter rather than as an absorption rate. A galvanic link injecting milliamperes is operating in that regime, which is precisely why published guidance for compliant galvanic operation constrains the injected current rather than the radiated power. A designer moving a body-coupled link from a few hundred kilohertz to twenty megahertz crosses from one exposure framework into the other, and the compliance argument must change accordingly.

Security and Privacy When the Node Is an Implant

IEEE 802.15.6 defines three security levels: level zero, which is unsecured; level one, which authenticates but does not encrypt; and level two, which authenticates and encrypts. It specifies several association procedures for establishing a shared secret, including ones that rely on a password or on an out-of-band confirmation such as a displayed value. Independent researchers have published cryptanalytic weaknesses in those association protocols, and the standard's security design should therefore not be adopted uncritically.

The deeper problem is that an implant inverts the usual threat model. In an ordinary embedded system, the asset is data and the worst outcome is disclosure. In a cardiac device or a drug pump, the asset is therapy and the worst outcome is a patient. Three consequences follow, and each of them cuts against the reflexes of conventional security engineering.

First, availability outranks confidentiality. An emergency clinician who encounters an unconscious patient with an unknown implant must be able to interrogate and, if necessary, reprogram it without the patient's cooperation and without a credential the patient carries. This break-glass problem has no clean cryptographic solution, so practical designs lean on physical proximity as an implicit credential: a magnet held against the chest, an inductive wand, or a near-field link that cannot be established from across a room. Proximity is a weak authenticator in the abstract and a strong one when the alternative is a locked-out patient.

Second, denial of service has a physical channel. An attacker who cannot decrypt anything can still keep an implant's receiver awake, and battery drain in an implant is not an inconvenience but a surgery. Wake-up radio design, rate limiting on unauthenticated requests, and hard caps on how much energy an unauthenticated party can cause the device to spend are therefore security controls, not power-management features.

Third, the update path is itself an attack surface and a liability. The canonical incident remains the FDA safety communication issued in August 2017 covering several St. Jude Medical, now Abbott, pacemaker families, whose remedy was a firmware update administered in a clinic. That episode established both that implanted devices ship with exploitable firmware and that patching them is a clinical procedure with its own risks. United States law has since caught up: section 524B of the Federal Food, Drug, and Cosmetic Act, added in 2023, requires sponsors of cyber devices to submit a plan for monitoring and addressing vulnerabilities, to make updates and patches available, and to provide a software bill of materials.

Privacy raises a separate class of problem that has little to do with cryptography. A device that advertises itself over Bluetooth Low Energy with a stable address is a tracking beacon, and advertising data that identifies a manufacturer of insulin pumps discloses a medical condition to anyone with a phone. Resolvable private addresses exist for exactly this reason, and using them correctly means rotating at a sensible interval and avoiding identifiers elsewhere in the payload that defeat the rotation. Beyond the radio, the data crosses into a phone, a cloud service, and often a clinical record, where it becomes special-category personal data under the European General Data Protection Regulation and protected health information under United States rules.

Body-coupled communication offers one genuine security property that radio does not: the signal is largely confined to the person, so an attacker cannot passively intercept a well-designed link from across a room. This is the property Microchip advertises for its BodyCom technology, in which a base unit's capacitive pad transmits a 125 kilohertz signal through the body to a tag that replies at 8 megahertz with an encrypted identifier, for applications such as door locks, vehicle access, and tool authorization. Touch becomes the authentication gesture. The property is real but not absolute; some energy always escapes, and confinement is defense in depth rather than a substitute for cryptography.

What Ships Today and What Remains Research

A clear-eyed account of deployment matters here, because the gap between the literature and the market is unusually wide in this field.

Deployed

Bluetooth Low Energy carries the overwhelming majority of real body area network traffic. Continuous glucose monitors, activity trackers, blood-pressure cuffs, pulse oximeters, and connected hearing aids nearly all use it, for one decisive reason: every phone already contains a compatible radio, so the product needs no dongle, no hub, and no ecosystem. No competing technology has overcome that advantage, and technical merit has not been sufficient to do so.

Implant telemetry to an external reader is mature. Cardiac rhythm devices from the major manufacturers communicate with bedside monitors and clinic programmers, and remote follow-up over such links is routine clinical practice. Legacy inductive telemetry at low frequencies has largely given way to MedRadio-band links, with some recent devices adding Bluetooth Low Energy specifically so that patients can use a phone application instead of a dedicated bedside unit.

Near-field magnetic induction is deployed in hearing aids for the ear-to-ear link between left and right devices, where a low-frequency magnetic link passes around the head more gracefully than a 2.4 gigahertz radio and consumes very little power. Newer devices carry a 2.4 gigahertz radio as well, for direct streaming from phones.

Conducted intrabody communication reached the market in a clinically significant device with Abbott's Aveir dual-chamber leadless pacemaker system, approved by the FDA in July 2023. Its i2i link exchanges subthreshold conducted pulses between the atrial and ventricular devices through the blood, on every beat, to maintain atrioventricular synchrony. The published Aveir DR i2i study reported successful de novo dual-chamber implantation in 446 of 452 patients, or about 99 percent, performed by 126 physicians, with twelve-month results meeting the study's safety and efficacy goals. This is the strongest existing evidence that galvanic-style intrabody signaling is a production technology rather than a laboratory curiosity.

Capacitive body coupling ships in access control and authentication products, of which Microchip's BodyCom is the best-documented example. The volumes are modest compared with radio-frequency identification, but the technology is real, purchasable, and in the field.

Research and Unrealized

IEEE 802.15.6 itself has seen very little commercial silicon. It is a carefully constructed standard that arrived after the market had already settled on Bluetooth Low Energy for on-body traffic and proprietary MedRadio links for implants, and standardization does not reverse an installed base. Its lasting contributions are its channel models, its coexistence requirements, and its explicit link-budget framework for radios operating against tissue, all of which continue to inform designs that do not implement the standard. The revision work in TG6ma may improve the specification, but adoption remains the open question rather than the technical content.

High-rate, ultra-low-power human body communication transceivers remain an active research area. Work on electro-quasistatic signaling has demonstrated links with very low radiated leakage and very low energy per bit, which if productized would strengthen the security argument considerably. Broadband body-channel characterization, measurement methodology that survives the grounding problem described earlier, and reliable channel models across postures and body types are all incomplete.

Batteryless implants powered entirely by harvested or beamed energy, ultrasonic links to deep implants, and very small distributed sensing nodes remain research programs with encouraging demonstrations and no routine clinical deployment. The same is true of most closed-loop bioelectronic therapies that depend on a fast, reliable, implant-to-implant link, which is precisely the application that would justify a dedicated body area network standard if any application would.

Conclusion

Body area networks are defined by their medium rather than by their range. The body absorbs, shadows, detunes, moves, and constrains exposure, and every architectural choice in this field traces back to one of those five facts. IEEE 802.15.6 organized the design space by defining a narrowband layer for conventional low-rate links, an ultra-wideband layer for higher rates and ranging, and a human body communication layer at 21 megahertz for signaling through the body itself, all under one medium access layer that trades power against latency. The spectrum follows the same logic: MedRadio and its 402 to 405 MHz core for implants under a 25 microwatt ceiling, MBAN spectrum at 2360 to 2400 MHz for hospital sensors, the Wireless Medical Telemetry Service for protected clinical telemetry, and unlicensed spectrum for its unmatched installed base.

Body-coupled communication works because tissue at these frequencies is a lossy conductor rather than a radiator, and because the resulting quasi-static coupling is far more efficient than any electrically tiny antenna would be. Galvanic coupling keeps both paths inside the body and buys robustness at the cost of range. Capacitive coupling achieves lower path loss and higher rates but depends on a return path through the environment measured in picofarads, which is why the usable channel sits in the megahertz rather than the kilohertz range and why measurement conditions must always accompany a published path-loss figure.

The engineering that follows is dominated by energy. An implant that must average microamperes for a decade cannot afford a radio that behaves like a phone, which is why conducted signaling, aggressive duty cycling, and architectures that avoid expensive frequency synthesis keep reappearing. Antennas must be tuned with tissue present and accepted as inefficient when implanted. Coexistence in the 2.4 gigahertz band is worsened by body shadowing and is now a documented regulatory deliverable through AAMI TIR69 and ANSI C63.27. Exposure limits bind mainly at the off-body tier and in wireless power transfer, while implants are more often limited by local heating, and low-frequency body-coupled links fall under a stimulation-based framework rather than a thermal one. Security must start from the premise that the asset is therapy rather than data, that an emergency clinician must be able to get in, and that battery drain is an attack.

The deployment picture should temper any enthusiasm drawn from the literature. Bluetooth Low Energy won the on-body tier because it is already in every phone, and implants use proprietary MedRadio links because they were validated before any standard existed. Conducted intrabody communication has nonetheless reached the market in a dual-chamber leadless pacemaker, and capacitive body coupling ships in access control, so the physics is not merely academic. The field's most useful posture is the engineering one: know which tier a link belongs to, know which constraint actually binds, and choose the physics that fits.

Related Topics