Cold Plate Leak Testing: Pressure Decay vs Helium Detection

A liquid cooling cold plate is only as good as its welds, and the test that proves those welds is the leak test. Two methods dominate production floors: pressure decay, which watches for pressure loss inside the sealed plate, and helium detection, which hunts for helium atoms escaping through it. They are not competitors so much as different tools - one cheap and fast, the other extraordinarily sensitive and correspondingly slower. Choosing between them (or deciding how to use both) is one of the defining quality decisions in cold plate manufacturing.

Pressure Decay: The Workhorse

The plate is pressurised - with dry air or nitrogen - to a set test pressure, sealed off, and the internal pressure is watched over a fixed time. A sound plate holds; a leaky one loses pressure at a rate the instrument converts into an estimated leak rate. The equipment is inexpensive, the test is quick, and the interpretation is intuitive, which is why pressure decay is the default screening test on most production lines.

Its limits are equally clear. Sensitivity is bounded by two things physics does not forgive: the temperature stability of the room (a one-degree drift changes the gas volume and mimics a leak) and the elasticity of the tooling and seals (bladder fixtures relax during the test and look like leakage). In practice, a well-built pressure decay station resolves leaks down to roughly the 10⁻³ to 10⁻⁴ mbar·L/s range - more than enough for many cold plate specifications, and not enough for the demanding ones.

Helium Detection: The Microscope

Helium leak testing uses helium - a small, inert, atmospherically rare gas - as the tracer. Two configurations matter in production. In vacuum-box testing the plate is pressurised with helium inside a chamber that is evacuated; any helium escaping is measured directly by a mass spectrometer, and the result is an absolute leak rate down to the 10⁻⁶ mbar·L/s range and beyond. In sniffer mode the plate is pressurised with helium and a probe is run along the welds and ports, locating where helium escapes. Vacuum mode quantifies; sniffer mode locates.

The costs are the flip side of the sensitivity: helium is a consumable that must be recovered or released, the mass spectrometer needs maintenance and a trained operator, and cycle times are minutes rather than seconds. The upside is equally stark: helium finds pinholes, micro-cracks and porosity that a pressure decay test would wave through - and it finds them before a customer does.

Sensitivity, Cycle Time and Cost per Part

Aspect	Pressure decay	Helium (vacuum / sniffer)
Practical sensitivity	~10⁻³-10⁻⁴ mbar·L/s	~10⁻⁶ mbar·L/s and below
Typical cycle time	1-5 minutes	5-20 minutes (sniffer: per weld)
Equipment cost	Low	High (plus helium consumption)
Operator skill	Basic	Trained
Best role	100% screening on every plate	Qualification, sampling, and 100% on critical plates

Because of that economics table, the common mature pattern is a two-stage gate: every plate passes pressure decay at end of line, and a defined sample - first articles, first piece after maintenance, and a fixed audit rate - goes to the helium rig. Critical products (AI server cold plates, laser cooling, anything where a leak destroys a far more expensive system) justify helium on every unit.

Setting the Acceptance Threshold

A leak rate limit should come from the working condition, not from the test equipment’s brochure. Work it backwards: the operating pressure and the coolant, the consequence of losing a given percentage of flow over the service life, and the detection margin the specification needs to catch a defect that is just becoming a leak rather than already failing. A plate running at 2 bar with water needs a far less aggressive threshold than a two-phase plate running near vacuum - and quoting the threshold in mbar·L/s on the drawing lets the test engineer pick the method instead of guessing what the designer meant.

Fixturing: The Quiet Variable

More false leak results come from tooling than from welding. Soft seals relax; quick-connect fittings weep; clamping that twists a thin plate opens and closes micro-channels at the ports. A pressure decay station should be validated with a known-good plate and a calibrated reference leak, and the blank-off test - pressurise the fixture with no plate at all - should be part of the daily check. If the fixture leaks, the process learns to blame welds that were innocent.

Feeding the Result Back to the Weld

A leak test is only useful if it says something about the welds. The two methods fail differently, and each points at its own defect family. Pressure-decay-relevant leaks - a macro porosity cluster, an undercut channel, a cold joint at a port - are usually visible on a cross section once the leak rate localises them. Helium-detected micro-leaks more often trace to fine porosity in the weld root or a hairline crack at the weld toe, which is a parameter story: too much power, wrong speed, or a fit-up gap that was marginal. When leak data is recorded per weld location rather than just per plate, the pattern tells the welding engineer exactly which joint design to fix.

Case Reference

A cold plate maker shipping AI-server plates screened everything with pressure decay and field-failed within months. A helium audit of returned units localized the failures to root porosity at two port welds - a joint the pressure decay threshold was never going to catch. Adding sniffer-mode helium at 100% on port welds and re-qualifying the port welding parameters (lower power, revised speed) brought field returns to zero within two quarters.

Conclusion

Pressure decay tells you fast whether a plate is obviously sound; helium tells you precisely whether it is sound at all. Use pressure decay as the screen, helium as the microscope, set the threshold from the working condition rather than the instrument, and feed the leak locations back into weld parameters. That loop - not any single instrument - is what keeps cold plates dry for their service life.

Cold Plate Leak Testing: Pressure Decay Vs Helium Detection