Contact Us
  • Home
  • BLOG
  • Why Does Passive Intermodulation Rise at a Plated Joint?

Why Does Passive Intermodulation Rise at a Plated Joint?

Passive intermodulation rises at a plated joint because the joint stops behaving like a resistor. Two strong carriers cross a boundary that holds a thin oxide film, a nickel layer, a loose contact or a metal particle, and that boundary mixes them the way a very weak mixer would. Plating is the largest single lever on the board: a nickel underplate below a thin gold contact typically costs 20 to 40 dB of PIM, and on a center contact the measured penalty has been reported at 40 to 50 dB.

The test that decides the number uses two tones at 43 dBm each, which is 20 W per carrier. A common specification is minus 150 dBc at that level, and minus 150 dBc below 43 dBm is an absolute product level of minus 107 dBm. The receive sensitivity of a wideband base station sits at about the same figure. The budget for the joint is therefore zero: a part that meets the number exactly sits on the noise floor, and anything worse eats into the uplink.

Why a plated joint behaves like a mixer

A metal to metal contact does not carry current across its whole apparent area. It carries it across microscopic asperities, and the rest of the interface is a gap of a few nanometres filled with oxide, sulfide or adsorbed water. Electrons cross that gap by tunnelling, or the film breaks down in tiny arcs, and either way the current voltage curve is not a straight line. A curved transfer function multiplies the two carriers, so products appear at 2f1 minus f2 and at 2f2 minus f1. In a frequency division duplex system the transmit carriers and the receive band are separated, and those products land inside the receive band.

The slope is what makes the effect sudden. Third order products fall 3 dB for every 1 dB the carriers drop, so a joint that is marginal at 43 dBm permfortable at 40 dBm and hopeless at 46 dBm. Running the carriers 3 dB hotter lifts the product 9 dB, from minus 107 to minus 98 dBm, which is 9 dB above the noise floor and no longer a margin question.

What the plating costs in decibels

The material in the current path decides most of the result, and the numbers are large enough to plan around. A silver plated N type connector has been measured at least 30 dB better than a standard part with a nickel or stainless steel outer contact, with the geometry unchanged. A stainless steel body alone adds 10 to 20 dB. A loose connector varied by about 40 dB over time, while the same part tightened to its rated torque held a constant level. Correct torque is not a handling detail, it is a PIM specification.

What sits in the current pathTypical penaltyWhat to do instead
Nickel under thin gold on the contact20 to 40 dB, up to 50 dB on a center contactSilver or gold over a non magnetic barrier
Stainless steel body in the RF return10 to 20 dBBrass or copper alloy body, plated
Loose or over tightened jointUp to 40 dB of driftRated torque, re checked after thermal cycling
Oxide film from residue and humidityIntermittent, bursts of a few dB to tens of dBResidue control and a cleaning step when the flux class needs it

Gold over nickel on a board pad repeats the same problem at a smaller scale, and the arithmetic is easy to check. SUPERB ENIG is 100 to 200 micro inches of nickel with 2 to 3 micro inches of gold, which is about 0.05 to 0.08 um of gold. At 2 GHz the skin depth in gold is close to 1.8 um, and 98 percent of the current density sits within about 4.6 skin depths, so roughly 8 um of gold would be needed to carry it. A 0.07 um layer is not a conductor at that frequency. It is a cosmetic and corrosion layer, and the RF current runs in the nickel beneath it, which is exactly the metal the PIM table warns about. A porous or worn gold layer raises the product without changing how the pad looks.

Where it starts on a board

The board level sources are few and they are all visible under magnification. The first is the ground tab of the connector: a joint with a void, or one that never reached full liquidus, interrupts the return current and adds a nonlinear boundary. The second is debris, a particle of machining swarf or a solder splash of 0.1 mm sitting across an RF pad, which touches and lifts as the assembly warms. The third is residue: a no-clean flux left on a high impedance node absorbs moisture, and the corrosion that follows is a semiconductor film.

Three habits that hide the problem

Proving it at low power. A two tone run at 33 dBm per carrier is a 10 dB lower stress, and the third order product drops 30 dB with it. A joint that fails by 10 dB in service passes this test with room to spare. The test level has to be the service level, which is why 43 dBm per carrier became the reference.

Believing gold means linear. Gold is a good conductor and a poor answer on its own. What matters is what sits under it and how thick the gold is.

Torque by feel and never again. A connector that is under torqued at build, or that relaxes after a hundred thermal cycles, drifts by tens of dB. Torque at the rated value, mark it, and re check after the environmental run rather than after the complaint.

What SUPERB controls on the build

PIM itself is measured with a two tone source in an RF laboratory, either yours or a partner laboratory, and we do not quote a PIM number that we did not measure. What we control is the hardware that decides it. Finish is specified per joint and recorded per lot: ENIG with nickel and gold thickness on the certificate, immersion silver, hard gold on contact edges, and immersion tin only where a tin surface is acceptable. Connector and shield joints are inspected by X-ray with void held under 5 percent to IPC-7095, and cleaning follows the flux class rather than habit.

The RF side of the build is on RF and microwave PCB, including mixed laminate builds and impedance held to plus or minus 5 percent, with plus or minus 2 percent on critical lines. Front end and module assembly is on RF modules, and the high layer count side is on RF and microwave board build.

What to send with the question

The carrier power and the specification in dBc or dBm, the frequency plan with transmit and receive bands, the connector family and its plating, the board finish with nickel and gold thickness, the torque value used at build, the cleaning class, and any field return that showed a rise after installation. With those seven items the answer is a plating and joint decision with a number behind it.

Need a quote or a review of your build against these limits? pcba@superb-tech.com · WhatsApp +86 133 9608 1443