When a 6.6 kW onboard charger fails out of warranty and the original Tier-1 has already moved its tooling to an 800 V platform, the only remaining asset is the physical board in your hand. Engineers who set out to reverse engineer onboard charger PCB assemblies are rarely chasing a competitor’s design; they are rebuilding manufacturable documentation for legacy hardware that still has years of service left, and for which no Gerber, netlist, or BOM survives. The task is harder on an OBC than on most automotive boards, because a single laminate carries mains-referenced switching nodes, a reinforced isolation barrier, and pack-side control logic that must never be bridged by accident.
Why an Onboard Charger Is Not Just Another Power Supply
An OBC performs, inside one die-cast housing, the job of an entire bench of instruments. AC enters through a fused inlet and a common-mode choke, passes an inrush limiter and a bridge rectifier, is corrected to near-unity power factor by a boost or totem-pole PFC stage, and lands on a 400 V DC link — or a 700–800 V link on newer architectures. From there an isolated DC-DC stage — usually an LLC, CLLC, phase-shifted full bridge, or dual active bridge — converts to pack voltage and provides the galvanic barrier between traction system and grid. An auxiliary flyback, a digital controller, a CAN link to the BMS and VCU, and control-pilot detection share the same board.
A mistake here is categorically different from one on a body control module. An ECU error produces a diagnostic trouble code; an OBC error produces an arc across a contaminated creepage gap, a welded charge relay, or a thermal event inside a pack. Insulation coordination follows IEC 60664-1, the vehicle interface IEC 61851-1 and IEC 62196, EMC is assessed against CISPR 25 and UN ECE R10, and every part in the power path is expected to be AEC-Q qualified. The environment is equally unforgiving: −40 °C cold starts, 105 °C ambient near the coolant plate, continuous vibration, salt spray, and a conformal coating that hides half the silkscreen.
Most of this work is compliant reverse engineering and repair support for legacy hardware: a charger family out of production, a fleet that still runs it, and no drawings left in anyone’s archive.
Four domains dominate such a design, and each one leaves a visible signature on the copper.
Power conversion. A bridgeless totem-pole PFC built on silicon carbide switches at roughly 65–100 kHz, with the current shunt in the return leg and Kelvin-connected sense traces running back to the controller. The DC link capacitor bank is the energy source that will still be charged when you pick the board up. On the isolated side, an LLC or CLLC stage runs at 100–200 kHz through a planar transformer that often integrates the resonant inductor into its own winding window, followed by synchronous rectification. The resonant tank values — magnetizing inductance, leakage inductance, resonant capacitance — set the gain curve and the zero-voltage-switching margin. A clone that matches the transformer’s footprint but not its magnetics will run, heat up, and lose efficiency at exactly the operating points that matter.
Signal integrity around switching nodes. Silicon carbide transitions at 50–100 V/ns, five to ten times faster than the silicon it replaced. Gate loop inductance has to stay in the single-digit nanohenry range, the source return must not be shared with the power stage, and desaturation blanking has to be tuned to the switch rather than inherited from an older driver. This is the part of an OBC layout that looks arbitrary in a photograph and is anything but, as our notes on MOSFET gate drive layout explain.
EMC. Y-capacitor placement decides where common-mode current returns, and the X-capacitor discharge resistor decides how fast the input is safe to touch. Choke winding style, shield termination, and the copper pour under the isolation barrier all show up in a CISPR 25 scan.
Safety distances. Reinforced isolation at 800 V means creepage in the region of 8 mm, frequently enforced by milled slots through the laminate, plus safety-rated Y-caps straddling the barrier. Those slots are geometry, not decoration, and they must survive into the Gerber data intact.
Teardown and Imaging a High-Voltage Assembly
The first rule is procedural: treat every OBC as energized until a rated meter says otherwise. A DC link can hold 400–800 V for minutes, and bleeders are sometimes undersized by design. Discharge through a proper resistor with insulated leads, never a screwdriver, then re-measure — dielectric absorption brings some of it back. Probe with the circuit unpowered and clipped.
Documentation comes before disassembly. Photograph both sides under even, diffuse light at high resolution; record connector pinouts, harness colours, screw torque, thermal interface material type and thickness, and the stack-up order of any pressure plate or cold-plate interface. Those details are trivially destroyed and impossible to reconstruct later.
Conformal coating and potting come next. Silicone and acrylic coatings yield to the right solvent; polyurethane and epoxy potting usually require controlled mechanical removal with the board fixtured, and every pass risks lifting a solder mask. Imaging follows: optical scanning for outer layers, X-ray for inner-layer copper, buried and blind vias, and planar transformer windings. Layer separation then proceeds one copper layer at a time, with registration held to about ±1 mil across six to ten layers.
Two measurements are routinely skipped and should not be. Copper weight per layer, taken from a microsection, because a 3 or 4 oz inner layer etched with 1 oz compensation produces traces thirty percent narrower than the original and will not carry the current. And slot geometry, because a milled creepage slot modelled as a routed outline is a different board from one where it was forgotten.
Net and Component Identification Across the Isolation Barrier
Trace from the barrier outward rather than from the connectors inward: establish the primary and secondary domains using the transformer pins, the optocouplers or digital isolators, the Y-cap return, and the physical slots. Only then does the rest of the board become readable.
On the primary side, expect an input protection chain of fuse, MOV, and X2 capacitor; an inrush network built from an NTC or a resistor bypassed by a relay; the PFC inductor and its shunt; the DC-link bank; and the controller with its bus-voltage divider. Faults cluster here — a welded relay contact after a bad inrush event, a shorted PFC switch, a dried electrolytic — a signature common enough that we document it in our work on power supply board reverse engineering.
Across the barrier, identify each coupler by function rather than part number. Gate drive for the secondary switches, voltage and current feedback, fault and enable signalling, and CAN isolation all cross the same boundary through different devices with very different failure modes. An optocoupler’s current transfer ratio degrades with age; a digital isolator usually fails outright. Distinguishing the two changes the diagnosis and the replacement strategy.
Component identification is where an OBC punishes optimism. Laser-abraded markings, potted daughter assemblies, and house-numbered controllers are normal. An LCR meter with a Kelvin clip, a curve tracer for junction and body-diode behaviour, and selective decapping resolve most of it; whatever remains goes on an explicit unresolved list instead of into a guess. Mark each net as confirmed, inferred, or unknown — a schematic that pretends to certainty is more dangerous than one that admits a gap. The same discipline applies to the pack-side interface; where the battery side is the actual subject, the approach overlaps heavily with EV battery management system board cloning.
How We Reverse Engineer Onboard Charger PCB Builds and Prove Them on the Bench
Documentation is only half the deliverable: a schematic, netlist, annotated board images, connector pin table, and BOM are useful only if a board can be built from them and made to behave.
The BOM is the first place a clone diverges from the original. Safety-critical parts cannot be substituted on ratings alone: Y-capacitors need the correct safety class, fuses the right breaking capacity, X-caps the right pulse rating. Power semiconductors need matching gate charge, output charge, and reverse-recovery behaviour, not just voltage and current, or the driver you inherited will be wrong for the part. Obsolete controllers are the usual long pole — the practical answer is a functionally equivalent device with an adapted compensation network, validated against the original’s measured loop response.
Prototyping then proceeds in two steps. A quick-turn board proves connectivity and basic function; only after that does the production stack-up follow, with the measured copper weights, the correct laminate, and the creepage slots cut exactly as the original had them.
Validation is where the project earns its credibility. Insulation resistance and a dielectric withstand test across the barrier, at a level derived from the working voltage rather than a number picked for convenience. Functional sweeps at low line, nominal, and high line. Power factor and input current THD across load. Resonant gain and efficiency mapping, because a tank that is slightly off shows up as heat, not as a dead board. A thermal soak at rated power on the intended cold plate, CAN communication against a BMS simulator including fault frames, and an EMC pre-scan, since a layout that shifted two millimetres near the Y-cap return can fail where the original passed. Our PCB EMC design reconstruction notes cover the layout decisions that most often cause that shift.
Note what this does not do: it does not certify anything. A reproduced board is a starting point for the customer’s compliance programme, and anyone promising otherwise is selling a shortcut that will not survive an audit.
Common Pitfalls That Sink OBC Clone Projects
Treating copper weight as a detail. The single most common cause of a clone that runs hot and fails early. Measure it, do not assume 1 oz.
Losing the creepage slots. Slots omitted from the rout layer turn a passing hi-pot into a flashover.
Copying the transformer by outline. Turns ratio is easy; magnetizing and leakage inductance are what the resonant tank actually needs. Measure both on an LCR bridge before committing to a winding.
Assuming firmware is recoverable. Read-out protection is normal on automotive controllers. Plan for a functional equivalent implementation or a donor module, and scope that work before quoting.
Validating only at nominal. A board that is clean at 400 V and 25 °C can be marginal at 800 V and −40 °C, where the PFC’s switching behaviour and the driver’s blanking time both move.
Ignoring the mechanical interface. Thermal pad thickness, mounting torque, and cold-plate flatness determine junction temperature more than copper does. Record them during teardown.
Boards that mix a mains-referenced PFC with an automotive control domain inherit two supply cultures at once; the AC-side grounding review is covered in AC-DC power supply board cloning.
FAQ: Practical Questions About Onboard Charger Reverse Engineering
Can an OBC be reverse engineered if the transformer is potted and the MCU is read-protected?
Yes, with a caveat on effort. A potted planar transformer cannot be unwound non-destructively, but its turns ratio, magnetizing inductance, leakage inductance, and interwinding capacitance can all be measured, which is enough to specify a replacement. Read-protected firmware is a different problem: the hardware can be fully documented, while the control behaviour has to be reimplemented or a donor module used. That boundary should be clear before the project starts.
Does a reproduced board need to be re-certified?
In practice, yes. A physically identical board is still a new assembly from a compliance standpoint, and the customer’s own safety and EMC programme has to cover it. We deliver documentation and prototypes that support that programme; we do not issue certifications and we do not describe a clone as certified.
What do you need in order to quote an OBC reverse engineering project?
The physical board, ideally two samples so measurements can be cross-checked, photographs of the assembly in place, any surviving documentation, and the intended use — repair support, a second source, or a redesign. Layer count, potting, and whether the controller is protected drive most of the scope.
Can bidirectional or 800 V designs be handled?
Bidirectional chargers add a second control domain and a symmetric CLLC or DAB stage, changing the gate-drive and sensing topology but not the method. Higher bus voltage raises creepage, clearance, and test-voltage requirements rather than tracing difficulty. Both are scoped case by case.
Getting the Documentation Back
An OBC that is out of production but still in service is an asset with no drawings, and that is a solvable problem. The work is unglamorous — discharge, photograph, measure copper, measure magnetics, trace from the barrier outward, and validate at the corners rather than the centre — but it produces something no photograph of a board can: a buildable, testable set of design data. If you need to reverse engineer onboard charger PCB hardware for a discontinued platform, send the board details through our contact form page, or review the wider scope of our PCB reverse engineering services first. We will tell you what is recoverable, what is not, and where the real risk sits.
