Wireless EV Charging PCB: How Heavy Copper, Parasitic Inductance, Gate Loops, and Foreign Object Detection Shape Board Design from 85kHz to 6.78MHz

Foreign Object Detection Thermal Shock, Delamination and Board Splitting

In my years of working on wireless charging PCBs, my biggest feeling is that many people in the industry put their eyes entirely on topology and switching frequency, but hardly discuss whether the board itself can withstand the heat. On the power loop of a Wireless EV Charging PCB, the current peak is routinely over a hundred amps, and especially when foreign object detection suddenly triggers a power reduction, the transient temperature rise on the copper foil can burn ordinary FR-4 to a different color. Once, in a system-level temperature rise test, a metal bottle cap was mistakenly placed at the edge of the coil, and the system pulled the transmitter current from 120A to almost zero within 300 milliseconds and then re-swept the frequency to try to recover. In the course of this sharp current collapse and recovery, the temperature of a 2 ounce inner copper sheet directly under the coil shot from 85°C to over 210°C, the board surface bulged instantly, and with the glass transition temperature of FR-4 only around 130°C, this kind of thermal shock simply cannot be withstood.

If you go to some multilayer pcb suppliers, they only produce a scheme based on conventional thermal simulation and do not understand at all how the local hot spots brought by magnetic field coupling should be absorbed in the stackup. For example, the magnetic field under the coil induces eddy currents on the inner-layer reference ground copper, with eddy current density highest in the projection area directly facing the coil conductors and decaying quickly at the edge, so the heating there is completely non-uniform.

Unfortunately many board factories’ simulations do not even set the anisotropic thermal conductivity of copper foil correctly, let alone couple in the hysteresis heating of the magnetic core and shielding layers.

A truly useful heavy copper pcb supplier will open their mouth and ask whether the inner copper directly under the coil should reach 4 ounces or more, instead of pushing a pile of expensive high-frequency board materials and calling it a day. They understand that cooling relies on heat capacity and thermal diffusion, and the thermal time constant of 4oz copper is more than twice that of 2oz copper, enough to press down the transient peak temperature.

But I also tried several heavy copper pcb manufacturers, some of whom clearly could not control resin fill when laminating thick copper boards, so the multilayer boards delaminated directly after a power burn-in and the resistance drifted to an unwatchable level. Once a supplier used a prepreg of high resin flow to laminate two layers of 6 ounce copper; at first glance the laminated thickness looked uniform, but after 48 hours of high-temperature burn-in, because the roughening treatment of the copper surface was insufficient, the bonding force between resin and copper foil decayed, and the Z-axis thermal expansion directly pulled off the roots of the inner-layer connection holes. The on-resistance drifted from 0.2 milliohm to over 10 milliohm, the efficiency of the whole power loop dropped a notch, and it took two whole weeks to pin down that it was PCB delamination.

Later I simply split the power loop and the signal loop onto two boards: the signal board went to a multilayer pcb manufacturer specializing in high-density interconnect, using laser blind vias and any-layer interconnect to keep digital and small analog signal routing clean; the power board went separately to a thick copper factory, with all inner layers required to be 4 to 6 ounces and with black oxide treatment of the copper surface mandatory to enhance bonding force, and the two boards were connected with high-current pin headers or soldered copper bars, which saved a lot of worry at once.

As for foreign object detection, do not listen to device makers boasting about how sensitive their detection chips are; on a real vehicle the parasitic coupling of the PCB can push the detection threshold into a state of repeated false alarms. I measured on a four-layer board prototype that the dv/dt of the power copper was injected directly into the input of the detection loop through interboard coupling capacitance, forming a common-mode pulse of nearly 20 millivolts amplitude and 50 nanoseconds width, while the wake-up threshold of the detection chip was only a nominal 15 millivolts, so with the vehicle parked at the charging position doing nothing, the system frequently triggered the foreign object detection state and the whole charging session was interrupted.

Only after being burned did I learn my lesson: I hard-reserved a 3 millimeter isolation slot between the digital circuit and the power copper, together with several deeply buried shielding copper layers, and only then got the interference under control. Specifically, along the edge of the isolation slot I punched a via fence going down through to the inner-layer shielding copper, and the shielding copper layer used a crossed mesh with 0.1 millimeter line width, which both kept the impedance of the eddy current discharge path below 2 milliohms and did not affect magnetic field penetration. The PCB here is not simply a current-carrying tool; it is a skeleton integrating the power path and the thermal path, and if the skeleton is not built right, the best control algorithm afterward is just watering flowers in the mud.

Later I summarized a quick method for judging a board factory’s level: ask them directly whether they can provide cross-section reports on resin-fill uniformity for three or four layers of inner copper of 4 ounces or more, and whether they have measured how the dielectric loss factor between copper foils changes at different frequencies. Those who can answer are basically worth deeper cooperation, while those who only stress the dielectric constant of the board material have most likely never crawled around in the pit of wireless charging boards.

FOD Coil Accuracy and Phase Detection on Heavy Copper Boards

On foreign object detection for wireless charging, many system people tend to think the work on the board is simple. You assume that placing the coil on the transmitter and connecting an analog front end to sample the amplitude is the end of it, but when you really do the work you know where the pits are. Two years ago I cooperated with a Heavy copper PCB manufacturer on a 7kW wireless charging transmitter board, in which the coil section directly used thick copper, a four-layer board with 4 ounce copper on both the top and bottom layers, used for the transmitter coil and the detection coil. Only then did I realize that the winding of the coil cannot be designed independently; it must be considered together with the PCB routing, otherwise the sensitivity is miserable.

At that time, because the power loop current was large, I found a Heavy copper PCB supplier in Shenzhen that specialized in thick copper boards and promised copper thickness uniformity within ±10%. When the first batch of samples came back, we found that the induced voltage of the detection coils at different positions differed enormously, with the maximum and minimum differing by nearly 9V. Investigation later showed that it was not an asymmetric winding of the coil but insufficient etching accuracy of the thick copper board, which caused deviations in coil line width and an uneven inductance distribution. Only then did I understand that the so-called improvement of magnetic field uniformity cannot be achieved by optimizing the coil winding alone; you have to pin it down from the manufacturing accuracy of the multilayer board. Later I changed to a multilayer pcb manufacturer who could do 6-layer mixed thick copper lamination, putting the detection coils on inner layers, using 0.5 ounce precision traces for the phase detection circuit, and still using thick copper on the outer layers to carry the large current, and the problem finally improved. This experience made me completely stop believing the claim that pure simulation optimization can solve FOD; there are too many variables on the hardware implementation side.

Phase detection is much harder to serve than amplitude detection. Because you need to sample the amplitude and phase of multiple coils at the same time, the phase consistency among the analog front-end channels must be controlled within 1 degree, otherwise the joint decision is useless. We used an eight-channel synchronous sampling ADC, but the trace delays were different and phase matching was not done well in the PCB layout, so a 2 to 3 degree phase change caused by coil position offset was misjudged as a metal foreign object. Many teams who look only at algorithms would never think of this kind of problem; they assume the decision algorithm is poor, but actually the multilayer pcb supplier had a deviation in microstrip impedance control during processing, which gave a poor phase specification between channels. Later we were forced to calculate the trace lengths ourselves and hard-match the path length from the coil to the ADC front end, pressing the error to within 0.2mm, and only then did it stabilize.

Another easily overlooked item is power stage noise coupling. On a high-power Wireless EV Charging PCB, the interference radiated by the GaN switches when they work goes straight into the detection coil loop, especially the narrow MHz pulses, whose spectrum overlaps exactly with the FOD detection band. Shielding cans and isolated grounds are only remedies; the fundamental method is still physical isolation. We pulled the detection coil and the receiving coil’s power loop apart on the board, with the ground layer of the multilayer board between them for isolation, and added a tunable band-pass filter at the front of the detection circuit, whose center frequency the MCU adjusts dynamically according to the operating frequency. This filter could not use an integrated part and was built from discrete devices, and it is very sensitive to PCB layout.

Overall, in wireless charging foreign object detection, choosing the right Heavy copper PCB supplier and multilayer pcb supplier is only the first step. What really tests your skill is how you communicate process tolerances with the board factory and how you make physical compromises among the coil, phase, detection circuit and power stage later on. Without having stepped into the pits, it is hard to appreciate these details.

6.78MHz Harmonics Versus the FOD Chain

For the last two years I have been tinkering with an MHz-class in-vehicle wireless charging system and have stepped in many more pits than in the earlier tens-of-kHz projects. At first I thought it was nothing more than raising the frequency, making the coil smaller and improving efficiency, but PCB-level problems almost overturned the whole scheme. The biggest conflict was in foreign object detection (FOD). The threshold measured nicely in the lab with a signal source, but once the whole machine ran at full power, FOD reported randomly: either it did not recognize a coin, or it tripped protection even with no load. At first I suspected that the software filtering was not enough, but a spectrum analyzer later showed that the harmonics generated by the power loop switching at 6.78MHz went straight into the FOD detection chain, with high-order components even soaring to tens of MHz, completely overlapping with the resonance point of the FOD coil. Only then did I realize that PCB layout at high frequency is not at all a matter of “connecting the wires”.

We were forced to re-prototype the entire Wireless EV Charging PCB and went around a circle of Heavy copper PCB suppliers, because the current density of the transmitter coil and the power loop was so large that ordinary copper thickness could not take the temperature rise. But a Heavy copper PCB manufacturer can handle current carrying and has difficulty guaranteeing impedance control on the signal layers, especially with the FOD small-signal traces right next to the large-current commutation loop, where inductive and capacitive crosstalk are a complete mess. Later we worked deeply with a multilayer pcb manufacturer who could do both high-copper-thickness multilayer and high-density interconnect, trying to build the FOD coil into an inner layer, isolated with an independent reference ground layer and shielding copper, with power ground and signal ground joined at a single point. That multilayer pcb supplier repeatedly adjusted several versions of the stackup structure, pinned the consistency of dielectric thickness and pressed with several different base materials before bringing the common-mode noise coupling we required down to an acceptable level. Because even a difference of a few microns in dielectric thickness makes the resonant frequency of the FOD coil drift, and the instability of Dk and Df, once amplified at MHz, directly raises the noise floor of the detection chain so that the threshold cannot be set stably.

Many people think a high-frequency PCB means choosing the most expensive high-frequency board material and adding a few layers of shielding cans, but at 6.78MHz and high power, structural parasitic coupling is the gate of hell. If the FOD coil routing sits slightly close to the drain copper pour of a switch, or the return path of the gate drive loop is not enclosed tightly, high-order harmonics pour into the receive chain, and however you tune the software decision threshold it is all in vain. For our last version of the design, even the screw hole positions and the grounding path of the shielding can were ground over with the board factory for half a month, otherwise a slight change in inductance would waste the entire FOD calibration. Running a few hundred watts now, FOD can basically be held stable, but every time I think of those months I still feel that planning the PCB of a high-frequency high-power wireless charger according to distributed electromagnetic behavior from the start saves much more of your life than going to suppliers for remedies after problems appear.

Supplier Lottery, Resonant Capacitor Placement and EMC

When I used to tune wireless charging boards in the lab, what annoyed me most was not that efficiency would not go up but that every time I changed suppliers, the boards that came back were like opening blind boxes. Once, in a hurry to prototype a 6.78MHz transmit-receive board, I took the easy way and found a manufacturer that had only made ordinary multilayer boards, and as the copper thickness went up, the interlayer bonding force collapsed, and after high-low temperature cycling the board blistered directly. Only after that did I understand that for something like a Wireless EV Charging PCB, finding the right Heavy copper PCB manufacturer is much more important than getting the topology to run.

Many people think the benefit of MHz-band wireless charging is that the coil can be made small and magnetic integration is easy, which is true, but there is a hidden pit: high-frequency current crowds even harder in the copper foil. In EV scenarios the resonant current is routinely over a hundred amperes, and even if you use Litz wire for the coil, the current-carrying density in the resonant capacitor loop and the rectifier bridge section on the PCB is several times that of the kHz band. Ordinary 2oz copper thickness simply cannot take long-term heat accumulation, and local hot spots can bake FR-4 above its Tg, after which the dielectric constant drifts and the resonance point flies away. This is also why I increasingly rely on Heavy copper PCB suppliers: not because of process mysticism but because the cost of trial and error is too high. Once we did a 12-layer board at a multilayer pcb supplier, requiring 6oz inner copper and, after the outer layers were finished with immersion gold, copper pillar reinforcement, and their engineer said directly it could not be done, because their lamination line simply could not control the shrinkage of such a thick core. Later we changed to an old-brand multilayer pcb manufacturer that specialized in power and automotive boards, who gave three lamination cushioning schemes alone and passed the prototype on the first try. Anyone who has done it knows that feeling of relief.

Another underestimated point is the influence of interlayer layout of the PCB on the magnetic leakage inductance. Under MHz conditions, the trace inductance between the coil and the resonant capacitor has reached a level that cannot be ignored, and if you add two more vias, the soft-switching range of the LLC is offset. On an EV wireless charging module I once tried moving the resonant capacitor from the top layer directly to an inner layer referenced in the same layer as the coil, and the leakage inductance dropped by about 18%, with switching loss visibly coming down. But this kind of change places very high demands on the layer shift control of the multilayer pcb manufacturer, because if the alignment of the coil pattern and the inner-layer capacitor pads is slightly off, the equivalent series resistance jumps up. I have heard that some peers use any-layer interconnect and stepped copper thickness on 6.78MHz EV prototypes, embedding the resonant cavity completely in the board and even eliminating the external capacitors, shrinking the volume by multiples. However, for that kind of board, the multilayer board suppliers in China who can stably mass-produce it can be counted on two hands.

Mentioning EMC opens another blood-and-tears library. Near-field coupling in the MHz band is slippery: you think adding a shielding frame settles everything, but in fact common-mode noise runs along the DC bus to the vehicle ground, conducts to the antenna end and stirs the magnetic field uniformity into a mess. I once ran into a case where FOD could not pass the automotive consistency test at all, and after a month of investigation we found that the ringing of a piece of copper on the PCB coupled with the parasitic mode of the coil near 6.78MHz. We changed to another multilayer pcb supplier to redo the stackup, split the power ground and used multi-point grounding, and only then pressed it down. Teams that have verified EMC only in the kHz band and then dare to boast that MHz is no problem are not something I much believe.

In the end, the role of the PCB in EV wireless charging is no longer that of a simple “circuit board”. It is the crossing point of high-frequency power, signals and magnetic circuits. Many times, however beautifully the scheme is fixed, one unreliable board can delay the whole project by half a year. When I choose suppliers now, I do not look at the quotation and lead time first; I first ask one sentence: have you shipped MHz-class Heavy copper PCB in volume, and for in-vehicle wireless charging? Those who answer vaguely, even if they call themselves the leader of multilayer boards, I dare not jump in with. In this trade, being bold is not what gets things done; it is the intuition accumulated from every failed prototype.

Gate-Loop Tolerance, Distributed Coupling and Partner Selection

People who build electric vehicle wireless charging sooner or later hit a wall: it is not that the principle is so hard, but that you have to fight to the end with the PCB in the MHz band. Many people think that finding a multilayer board manufacturer to pile on a few layers of copper is enough, but once it comes to the gate drive loop, they are exposed immediately. A deviation of a few nanohenries is enough to make the switching waveform shake like an electrocardiogram, and the heat directly eats up the efficiency; this is no joke.

During the time I was looking for a Heavy copper PCB supplier, my biggest feeling was that factories that can do thick copper are a dime a dozen, but those who understand high-frequency coupling and thermal margin balance are few. What you want is not just to raise the copper thickness but to keep the power layer and the gate return layer close enough, with the dielectric thin enough that the magnetic field is tightly coupled, while still surviving thermal cycling. In this kind of multilayer board design, where to place the vias and which two layers are laminated tight against each other matter much more than merely choosing a heavy copper PCB manufacturer. Some factories stress how strong their current-carrying capacity is, but once put on a vector network analyzer, the parasitic parameters are a mess, and the gate voltage overshoot cannot be reined in at all.

Why is coupling at MHz so deadly? Because at low frequency you can still play with components as lumped parameters, but at several megahertz the whole board becomes a distributed system, and the little energy coupling between traces converts directly into EMI. The metal enclosure, the coil bracket and even the heat sink all take part in this field, and resonance points that do not exist in the kHz band all appear at MHz. At this point you rely on the traditional EMC way of testing, measuring conducted and radiated emissions, but without a near-field scan you are just guessing, yet many multilayer pcb suppliers in China do not even raise this with you; what they deliver is electrical connectivity, not field distribution.

The inductance of the gate loop is something with zero tolerance. A dual-switch topology looks like it can save gate drive channels, but once single-device stress goes up, the worst-case junction temperature calculation cannot allow the slightest wishful thinking. I once got careless in a thermal calculation, and at high temperature the gate drive waveform collapsed at the edge, the half bridge shot through, and the board burned in a miserable state. Afterward I would rather make the topology “dumber” and physically press the gate loop to the shortest, while forcing myself to choose multilayer pcb manufacturers who truly understand high-frequency electromagnetic and thermal coupling, who are willing to spend on thin prepreg in the stackup coupling and optimize the via anti-pads. These details are worth more than the numbers on the quotation.

A last word on the detection side. For foreign object detection, if the analog front end is not physically isolated from the high-power traces, MHz switching harmonics pour straight into the detection signal, and whatever FOD algorithm you use is in vain. You must separate the signal return paths completely and, at the PCB level, make their grounds free of common-mode interference. I also gained this understanding only after burning a pile of boards. So when I pick a wireless EV charging PCB partner now, I first look at their high-frequency cases and then at their production line capability; if they have not done actual near-field coupling measurements around 6.78MHz, I basically do not talk to them.