Chapter 5
Phase 1 of WIA-COMM-009 fixed what a wireless-power system looks like on the wire. Phase 2 is where a design team makes the physical decisions that the envelope later describes, and then exposes the result through an HTTPS and WebSocket surface. Those decisions come in a fixed order: requirements, technology class, coil or antenna, power electronics, safety integration, and test. This chapter follows that order. It then shows how the API makes the safety and regulatory consequences of those decisions enforceable at the moment a link is opened.
The specification's design process begins with five requirements, and every later choice is checked against them:
Power and distance together select the technology class. The specification gives this rule of thumb:
| Power | Distance | Technology |
|---|---|---|
| < 100 W | < 10 mm | Inductive (Qi) |
| 100 W – 50 kW | 10 mm – 2 m | Resonant |
| > 1 kW | > 2 m | Microwave / Laser |
The rows overlap above 1 kW, and that overlap is intentional. Distance is the discriminating variable. A 3 kW link across 5 mm is still a short-gap magnetic problem. A 3 kW link across 20 m cannot be solved with coils, because the coupling between two coils falls off rapidly once the gap is large compared with the coil diameter. Treat the table as a first filter, not a decision procedure.
Once the class is chosen, each discipline gets its own simulation tools. The specification names these:
On the electronics side the guidelines name full-bridge and half-bridge inverter topologies. They also name PLL-based frequency tracking, so the transmitter stays on resonance as coupling changes. For the in-band communication that lets receiver and transmitter negotiate power, they list ASK, FSK and Bluetooth.
The physics underneath Phase 2 is a set of trade-offs rather than a single optimum. Raising the operating frequency shrinks the coils but increases AC losses. The specification therefore records the frequencies that the main product families settled on, together with the reason for each:
| Family | Frequency | Rationale given |
|---|---|---|
| Qi | 80–300 kHz | Compromise between coil size and loss |
| Resonant | 6.78 MHz | ISM band, higher power density |
| EV charging | 85 kHz | Penetration and efficiency |
Efficiency work proceeds in three places: the coil, the matching network and the rectifier.
Coil. Above 20 kHz the guidelines call for litz wire, because skin effect pushes current to the conductor surface and a solid wire wastes its core. Where current is very high and cooling is the concern, copper tube takes over. The ferrite behind the winding is chosen by frequency: Mn-Zn below 1 MHz and Ni-Zn above it. The trade is permeability against losses, since a material that guides flux well usually dissipates more of it as heat.
Matching. Maximum power transfer requires the load impedance to be the complex conjugate of the source impedance:
Z_load = Z_source*In practice an L-C matching network transforms the real impedances toward that condition. As coupling shifts with alignment, the match drifts. This is the practical reason for the PLL and frequency tracking in the inverter stage.
Rectifier. The rectifier is where the received AC becomes usable DC, and every diode drop is paid for in heat. The guidelines compare three options:
At low output voltages a fixed diode drop is a large fraction of the total, which is why synchronous rectification pays off most there.
Thermal budget. Whatever efficiency is achieved, the remainder becomes heat in the transmitter and receiver:
P_loss = P_tx × (1 - η)
11 kW EV charger @ 90 % efficiency:
P_loss = 11,000 × 0.10 = 1,100 WThat 1,100 W must be removed from a sealed pad that sits on or under a road surface. The specification maps loss to cooling method:
| Dissipation | Cooling |
|---|---|
| < 50 W | Natural convection |
| 50–500 W | Forced air |
| > 500 W | Liquid cooling |
The worked EV example therefore lands in the liquid-cooling regime. A phone charger at a few watts never leaves natural convection. Efficiency points are worth more at high power, because each lost percentage point is multiplied by a larger transmitted power.
The specification's cost analysis looks at the EV Level 2 stationary charger rather than the market as a whole, since that is where a bill of materials is large enough for design decisions to matter.
| Component | Cost (per unit) | Scaling driver |
|---|---|---|
| TX coil (EV, 11 kW) | $200–500 | Volume production |
| RX coil (EV) | $300–600 | Integration |
| Power electronics | $500–1500 | Semiconductor cost |
| Control / comm | $100–300 | MCU, sensors |
| Installation (stationary) | $500–1000 | Labor |
| Total (EV Level 2) | $1,600–3,900 | Target: < $2000 |
The column sums are consistent with the stated total. The lower bounds add to $1,600 and the upper bounds to $3,900. The target of under $2,000 is reachable only near the bottom of every range at once. That makes the driver column the useful part of the table: only coil production volume and semiconductor cost are expected to fall through manufacturing scale. Installation labor tends not to. A designer who needs to hit the target therefore has to attack power electronics and coil integration first, because those are the lines where the specification expects scaling to help.
The technology classes in 5.1 map onto distinct product settings, and each setting stresses a different part of the Phase 2 guidance:
The specification places safety integration before testing in the design process, because retrofitting detection onto a finished design is rarely successful. It lists three safety elements: FOD and LOP sensors, thermal management, and EMI shielding. The test stage then verifies these against external references:
| Concern | Method or reference named |
|---|---|
| Efficiency | Power-analyzer measurement (Yokogawa WT5000) |
| Emissions (EMC) | CISPR 11, FCC Part 15 |
| Human exposure (SAR) | IEEE 1528, IEC 62209 |
Efficiency is measured, not computed, because the number that feeds the thermal formula in 5.2 must come from the real assembly. The measured value also appears later in the link telemetry. The specification's API turns the same concerns into runtime gates. A link may only open if the transmitter's safety state is currently valid, and the system's regulatory authorisation is a first-class record that integrators can query. Those mechanisms are covered next.
The Phase 2 surface is small. It is built around the lifecycle of a physical power link: register hardware, prove it is safe, open a link, watch the link, and answer to the regulator.
POST /wireless-power-transfer/v1/systems # register WPT system
GET /wireless-power-transfer/v1/systems/{id} # fetch system record
POST /wireless-power-transfer/v1/links # establish link
GET /wireless-power-transfer/v1/links/{id}/state# current link state
POST /wireless-power-transfer/v1/safety/test # run safety self-test
WS /wireless-power-transfer/v1/telemetry/stream# live link telemetry
GET /wireless-power-transfer/v1/regulatory/{id} # regulatory authorisation statusDiscovery follows the convention at /.well-known/wia-wireless-power-transfer.
Registration and links. POST /systems accepts the Phase 1 envelope and returns a stable systemId. POST /links then opens a link between a registered transmitter and a registered receiver. The request carries per-link parameters for alignment, requested power and link duration. The endpoint validates the request against the regulatory envelope, and the rules differ by class:
A coil-based link needs neither check. This is why 5.1 begins with the technology class: it determines which gates the API will apply.
Safety self-test. POST /safety/test runs four checks:
The endpoint returns a result per test. A failed result blocks later POST /links requests until the fault is fixed and the self-test passes again. In effect the test is a precondition for energising the link, not an advisory report. The power-meter check also protects the efficiency figures: a miscalibrated meter would corrupt both the thermal estimate and the telemetry.
Telemetry. The WebSocket multiplexes per-link events:
Subscribers filter by link class and by event class. The broker emits alarm events on safety-trigger conditions as defined in Phase 3 §A.6. The telemetry fields mirror the design quantities from earlier in this chapter. The efficiency η and the loss formula P_loss = P_tx × (1 − η) can be evaluated live, and the thermal status shows whether the cooling choice from 5.2 is holding.
Regulatory and audit. GET /regulatory/{id} returns the authorisation envelope: a per-jurisdiction licence reference, the validity period, any conditional restrictions, and the post-issuance compliance-monitoring envelope. The audit endpoints expose an immutable history of registrations, safety-test outcomes, link establishments, regulatory events and unscheduled shutdowns, with integrity anchored in a per-tenant Merkle tree.
Limits. The rate limits are deliberately asymmetric:
| Surface | Limit |
|---|---|
| System and link endpoints | 1000 req/h, authenticated |
| Safety self-test, per system | 60 req/h |
| Telemetry WebSocket | 50 simultaneous subscriptions per credential |
| Microwave / laser / space-solar regulatory endpoints | Additional per-jurisdiction quotas |
The tighter self-test limit exists to defeat abuse. Without it, a client could repeat the test until a marginal fault happened to pass.
For a first deployment, the specification suggests starting from the reference container, which serves mock data for every envelope and endpoint. Integrators run the conformance suite against it, then replace the mock backend with real hardware one endpoint at a time. Telemetry and audit replication come last.
The specification already anticipates the high-power, long-range end of the field by including microwave, laser and space-solar systems in the same API as phone chargers. What differs is the regulatory envelope, which is stricter for the beam-based classes. Within the 1.x line the endpoint set is fixed. Hosts may add optional fields and new envelopes, but may not remove existing ones, so integrations written now keep working as new classes mature. Breaking changes are reserved for a major version with a 12-month deprecation window.
Phase 2 turns a set of engineering trade-offs into a service contract. The design sequence starts from power, distance, frequency, efficiency and cost, and selects among inductive, resonant and beam-based classes. Frequency, litz wire, ferrite grade, matched impedance and rectifier technology then fix the efficiency. That efficiency feeds directly into P_loss = P_tx × (1 − η), which decides whether a pad can be air-cooled or needs liquid cooling. The 11 kW EV example loses 1,100 W at 90% efficiency, and its cost range of $1,600–3,900 explains why the under-$2,000 target depends on volume and semiconductor scaling.
The API then enforces what the design promised. Registration records the system, the self-test proves interlocks and calibrations are valid, link creation is refused when authorisation or test freshness is missing, and telemetry reports the same quantities the designer calculated. Rate limits and per-jurisdiction quotas keep the safety and regulatory endpoints from being used as a way around those gates.
What this chapter has not covered is behaviour across trust boundaries. The telemetry stream mentions alarm events defined in Phase 3 §A.6, and the regulatory data raises the question of how operators, authorities and peer hosts exchange it. Chapter 6 picks up there. It describes the protocol exchanges that carry these envelopes between parties, and what happens on the wire when a safety trigger fires.