RF PCB reverse engineering recovers the full manufacturing package from a physical microwave board — layer-by-layer copper images, drill data, a stackup with measured dielectric thicknesses, impedance targets, BOM and schematic — accurate enough that a rebuilt board behaves the same way in the band it was designed for. The hard part isn’t tracing nets. It’s layer separation and stackup measurement, because on an RF board the dielectric is a circuit element, not just insulation.
If you’re holding a board with no files and a 2.4 GHz front end, a GPS chain, or a 900 MHz PA on it, that’s the question you’re really asking: can anyone give me copper and stack data I can build to? Yes, but only if the measurements happen in the right order.
Why an RF board can’t be copied like a digital board
On a four-layer logic board, if you get the netlist right, the board works. Nobody cares whether the core under layer 1 is 0.20 mm or 0.25 mm. On an RF board, that 0.05 mm changes the width-to-height ratio of every microstrip and moves a nominal 50-ohm line by several ohms. Copy the copper and guess the dielectric, and you get a board that passes continuity and fails VSWR.
So an RF job is really three measurement tasks running in parallel:
- Copper geometry — not just connectivity, but the actual shape of every line, taper, coupled-line filter, open stub, pad radius, keep-out and ground-plane cut, captured as polygons rather than redrawn as traces from a netlist.
- The dielectric stack — thickness of every core and prepreg, base copper versus plated copper, solder mask coverage and thickness over and beside the RF lines, and which laminate family the material plausibly belongs to.
- Ground integrity — via fences alongside transmission lines, stitching pitch, thermal via arrays under the amplifier, coplanar-waveguide gap dimensions, and whether any vias are blind, buried or back-drilled.
Miss the via fence pitch and adjacent channels start talking to each other. Miss the surface finish and you change conductor loss — ENIG and immersion silver do not give the same insertion loss at 6 GHz, because the nickel underlayer is lossy and the skin depth up there is tiny. Miss a hybrid construction, where a thin high-frequency laminate is bonded to an FR-4 sub-stack for mechanical support, and you’ve thrown away the entire design intent of the board.
This is also why RF work costs more per square inch than a control board of the same size. You’re not buying tracing labor. You’re buying metrology.
How layer separation works in RF PCB reverse engineering
The order matters, because most of the useful measurements can only be taken while the board is still intact.
Non-destructive measurement first
The board gets photographed and scanned at high resolution on both sides, then examined under X-ray to reveal inner-layer copper, via structures and buried features before anything is cut. Overall thickness is measured with a micrometer at several points across the panel, because the middle and the edges are rarely identical.
If the RF launches are accessible — an edge-launch SMA, a probe pad, a connector footprint, a test coupon on the tab — we take TDR readings and, where a fixture can be built, network analyzer sweeps on the original board. Those numbers become the acceptance criteria for the rebuild. Most shops skip this step, and that’s exactly why their rebuilt boards can’t be shown to be equivalent to the original. A netlist match proves nothing about a filter skirt.
Component and marking survey
RF parts are the identification bottleneck. Amplifiers, mixers, switches and LNAs in small packages often carry a two- or three-character factory code instead of a full part number, and SAW or BAW filters are frequently house-marked with nothing but a lot code. Ceramic baluns and chip antennas usually carry no marking at all.
So every marking is recorded under the microsc
Working on a board like this?
Send the chip marking or two photos. You get feasibility, lead time and price within 24 hours, and the check costs nothing.
Get a free quote