Contact Us

Interconnect Stress Test

Interconnect Stress Test

A board leaves the line with every net tested, every impedance coupon inside tolerance, and a microsection that looks clean. Two hundred thermal cycles later one unit opens, and the failure analysis shows a crack at the knee of a buried via. No single measurement on that lot was out of specification. What was missing is a test that loads the whole interconnect at once, plated barrel plus capture pad plus laminate, and watches resistance while it happens.

An interconnect stress test does that. It treats the board as a chain of joints instead of a list of features, and it is the test buyers of high reliability boards ask for before a first article is released.

What the test measures

A coupon carries a daisy chain of vias, connected in series and brought out to two pads. Current is passed through the chain to heat it, the chamber holds the ambient at the second temperature, and a four wire measurement reads the resistance of the whole chain. The coupon is cycled between the two temperatures, and the resistance is logged at every cycle.

Failure is a resistance rise past the agreed limit, or an open. The result is not a pass mark on a feature, it is the number of cycles the interconnect survived, plus the cycle at which the rise started. That second number is the useful one, because a chain that begins to drift at cycle 80 and opens at cycle 400 tells a different story from one that holds flat and then breaks.

What the coupon has to contain

The coupon decides whether the test means anything. It has to be built on the same panel as the lot, with the same drill sizes, the same aspect ratio, the same plating thickness and the same lamination cycle. A coupon with a 10:1 via chain does not qualify a board built at 20:1, because the failure mode that matters at high aspect ratio only appears at high aspect ratio.

Most requests also ask for more than one chain: the smallest via, the deepest via, and a chain through the layer pair that carries the highest speed signal. Three chains on one coupon cost little and answer the question a single chain cannot.

Interconnect stress test, thermal cycling and thermal shock

TestWhat loads the jointWhat it provesWhere it fits
Interconnect stress testCurrent through the chain plus chamber air, resistance monitored liveCycle life of plated barrels and their capture padsLot release and process change checks on high reliability boards
Air to air thermal cyclingChamber air only, whole assemblyJoint life including components and solderAssembly level qualification, slower per cycle
Thermal shockLiquid to liquid transferSurvival of a severe gradientMilitary and space screening, small samples
Solder float and reflow simulationOne or three passes at reflow temperatureThat the board survives assemblyEvery lot, as a process check rather than a life test

The difference in practice is time and resolution. A chamber cycle takes an hour or more, so a 500 cycle programme runs for weeks. An interconnect stress test cycle runs in minutes because the current does part of the heating, so the same 500 cycles fit into days, with resistance recorded at every one of them instead of at the end.

The numbers that set pass or fail

Four values come from the drawing or from the purchase order, and they have to be written down before the first coupon goes in:

  1. Cycle count. Three hundred cycles is a common target for industrial and telecom boards, and higher counts appear on military and medical programmes. The count is the buyer's call, not the board house's.
  2. Temperature window. 25 to 150 degrees Celsius is the usual pair. A wider window is harsher per cycle, so a 300 cycle programme at a wider window is not the same test.
  3. Resistance limit. Ten percent rise over the initial reading is the figure most often quoted. The initial reading matters, because a chain that starts high has less headroom.
  4. Sample size and coupon location. One coupon per lot is a process check, three coupons per lot is a statistic. Say which you are buying.

If the drawing is silent on all four, the report will be written against whatever the lab uses by default, and comparing it with a second source later becomes guesswork.

How to read the report

Start with the initial resistance per chain. A chain that reads high before cycling usually means a thin barrel or a poor capture pad, and the cycling only confirms it. The barrel thickness that is normal for a given hole is covered under plated hole copper thickness. Then look at the delta curve: a step change at one cycle points to a single weak via, a steady ramp across all chains points to plating quality across the panel. Then read where the failure sat, layer pair and position, and check that the cross section taken after the test shows the crack in the barrel and not at the pad to barrel transition. The two are different process problems.

How SUPERB PCBA handles a request

Boards are built to print at the PCB end in Jiangsu and Zhuhai, with coupons from the production panel and not from a separate setup. The published envelope covers 1 to 20 layers on the FR-4 line and up to 68 layers on the high layer line, aspect ratio to 25:1, plated copper of 20 micrometres or more in the barrel, and back drill stub of 0.25 mm or less where a high speed layer pair needs it. Coupons are checked at our own bench for resistance and continuity, and the cycling itself runs on the coupon profile agreed with the customer, with the cross section and the raw resistance log shipped with the lot.

RF and high speed boards keep their own constraints inside the same flow, covered under RF and microwave PCB capability, and every lot can ship with the impedance and microsection records that sit next to the stress data.

Questions buyers ask

Does every lot need an interconnect stress test?

No. Most programmes run it on the first article, on a process change, and then at a set frequency such as quarterly. Every lot gets continuity, impedance and microsection instead.

Can it replace thermal cycling of the assembly?

No. It loads the bare board interconnect. Once components are on the board, joint life is a different question and belongs to assembly level cycling.

What if the coupon passes and a board still fails in the field?

Then the coupon did not represent the board. Check aspect ratio, drill size and layer pair first, because a coupon that differs on any of the three qualifies something else.

Next step

Send the Gerber set, the stack up and the four values above: cycles, temperature window, resistance limit and coupons per lot. The reply comes back with a coupon drawing for approval, the cycle profile, and a sample report so the acceptance columns are agreed before the first panel is built.