Random vibration and thermal cycling are the two stresses most production screens rely on, and they precipitate different defects. Learn which workmanship defects each stress finds, why, when to combine them, what each costs in time and equipment, and how to choose for a given product.
Two Stresses, Two Mechanisms
A screen works by applying a stress that turns a latent defect into a failure the test can detect. The two stresses do this through different physics.
| Random vibration | Thermal cycling | |
|---|---|---|
| Load on the defect | Inertial force from acceleration, amplified at each resonance | Strain from differential expansion between bonded materials |
| What drives it | Mass of the part, stiffness of its support, input spectrum | Difference in coefficient of thermal expansion (CTE), temperature range, part size |
| Typical exposure | Minutes per axis | Hours (several cycles, each with ramps and dwells) |
| Defect location | Where mass hangs on a weak attachment | Where two materials with different CTE are joined |
A defect sits in one of these columns more than the other. That is why a screen built on one stress alone leaves a predictable class of defects untouched.
What Vibration Finds
Under acceleration, every component pulls on its attachment with a force proportional to its mass (F = m × a). Light surface-mount parts on a stiff board see little load. Heavy parts, long leads and cables see much more, and the board itself bends at its resonances.
Defects that random vibration precipitates:
- Loose or under-torqued hardware: screws, standoffs, nuts on terminals, heatsink clamps
- Cracked or cold solder joints on heavy components: large electrolytic capacitors, inductors, transformers, through-hole connectors
- Joints on wires and cables: power leads soldered to pads, motor phase wires, battery leads, with or without strain relief
- Connectors: partially seated mating, missing latches, crimp contacts with too few strands
- Unsupported mass: components standing on their leads without adhesive, cable bundles without ties
- Particles and debris: solder balls, wire clippings, which move and short under vibration
A cold joint on a heavy wire can read a normal resistance at rest, because the cracked surfaces still touch. Under vibration the contact opens and closes, and the resistance fluctuates. A test that measures continuously during the vibration sees this. A test after the vibration often does not.
How Resonance Concentrates the Load
The board and its heavy parts respond most at their natural frequencies. Miles' equation estimates the RMS acceleration response of a single resonance excited by a flat random spectrum:
Grms_response = sqrt( (π / 2) × fn × Q × PSD )fn: natural frequency of the part or board, HzQ: amplification at resonance (dimensionless)PSD: input power spectral density at fn, g²/HzExample values: fn = 300 Hz, Q = 10, PSD = 0.04 g²/HzGrms_response = sqrt(1.571 × 300 × 10 × 0.04) = sqrt(188.5) ≈ 13.7 gThe example values are illustrative. The point is the dependency: the response grows with Q, so a heavy part on a flexible support concentrates the screen on exactly the attachment that a workmanship defect weakens.
What Thermal Cycling Finds
When temperature changes, each material expands at its own rate. Where two materials are bonded, the difference is taken by the joint between them as strain. For a solder joint between a component and a board, the shear strain scales as:
γ ≈ (Δα × ΔT × L) / hΔα: CTE difference between component and board, 1/KΔT: temperature swing, KL: distance from the component's neutral point to the joint, mh: solder joint height, mExample values: Δα = 10 ppm/K, ΔT = 125 K, L = 5 mm, h = 0.1 mmγ ≈ (10e-6 × 125 × 5e-3) / 1e-4 = 0.0625Again the numbers are an example. The dependencies are what matter: large packages (long L), thin joints (small h), and material pairs with a large CTE difference concentrate the strain. A joint with a void, insufficient wetting or an existing crack has less load-bearing area and fails first.
Defects that thermal cycling precipitates:
- Solder fatigue at marginal joints: voids, insufficient fillets, poor wetting, especially on large ceramic and BGA packages
- Plated through-holes and vias: barrel cracks from the board's expansion through its thickness
- Component defects: cracked ceramic capacitors, die attach voids, wire bond weaknesses
- Delamination: in the board, under conformal coating, in potting
- Seals and adhesives: leaks in sealed enclosures, adhesive bonds between mismatched materials
- Parametric drift: components that move out of tolerance at temperature extremes
Thermal cycling also exposes defects that are not mechanical: a marginal oscillator that fails to start cold, a timing margin that closes hot. These appear only if the unit is powered and monitored during the cycles.
Combined Profiles
Some defects need both stresses at once. A crack in a joint may stay closed at ambient under vibration, and stay closed under thermal cycling without vibration, but open when the board is cold (materials stiffer, solder less ductile) and vibrating. Combined screens apply vibration during the thermal cycle, usually during the cold and hot dwells.
| Approach | Equipment | Notes |
|---|---|---|
| Sequential (vibration, then thermal, or the reverse) | Separate shaker and chamber | Simplest; each stress tuned independently; misses defects that need both at once |
| Combined in one chamber | Chamber with an integrated vibration table | Catches combined-stress defects; vibration and thermal parameters interact through the fixture |
Chambers used for HALT and HASS often combine a fast thermal chamber with a pneumatic repetitive-shock table. That table excites several axes at once with a broadband spectrum, which differs from the single-axis, controlled-spectrum excitation of an electrodynamic shaker. A profile developed on one type does not transfer directly to the other.
Cost and Time
The two stresses cost different things.
| Cost driver | Random vibration | Thermal cycling |
|---|---|---|
| Time per unit | Short: minutes per axis | Long: several cycles of ramp and dwell, set by the product's thermal mass |
| Units per load | Limited by the shaker table area and fixture | Many units per chamber load, if airflow reaches each one |
| Fixture | Product-specific fixture, designed and surveyed for resonances | Rack or tray, rarely product-specific |
| Monitoring | Cabling must survive vibration | Cabling must pass through the chamber wall and survive the temperature range |
| Profile risk | Fixture resonance can over-test or under-test the product | Dwell too short leaves the product short of the setpoint |
Thermal cycling time is dominated by how fast the product, not the air, changes temperature. Increasing the chamber ramp rate shortens the cycle only if the product follows. Measure the product with a thermocouple during profile development.
How to Choose
Start from evidence, not from habit. The table below maps observations to the stress that addresses them.
| Observation | Stress to apply first |
|---|---|
| Field returns with cracked joints on heavy parts, wires or connectors | Random vibration |
| Field returns after a number of temperature cycles in service, fatigue cracks on packages | Thermal cycling |
| Product has heavy components, hand-soldered wires, cable harnesses, screwed assemblies | Random vibration |
| Product has large ceramic or BGA packages, mixed materials, sealed or potted enclosures | Thermal cycling |
| Product sees vibration and temperature together in use (vehicles, aircraft, drones) | Both, consider combined |
| Defects that pass the functional test but fail intermittently in the field | Whichever stress reproduces the intermittent, with monitoring during the stress |
When neither the product nor the field data clearly points to one stress, run both for a period and record which one precipitates each failure. That data settles the question for the product.
Measure Which Stress Finds What
The way to decide on data is to record each stress as a separate phase, with the same monitoring measurement in each. With TofuPilot, a procedure with a vibration phase and a thermal cycling phase, each with a fault count measurement, stores per serial number which stress the unit failed under.
procedure.yaml40 lines
name: Controller Board Screenversion: 1.0.0execution: workers: 1 # the unit goes through one stress at a timeplugs: - name: Shaker Controller key: shaker python: plugs.shaker:ShakerController - name: Thermal Chamber key: chamber python: plugs.chamber:ThermalChamber - name: Unit Monitor key: monitor python: plugs.monitor:UnitMonitormain: - name: Random Vibration Z key: vibration_z python: phases.vibration:random_vibration with: axis: "Z" duration_s: 600 measurements: - name: Vibration Faults validators: - operator: "==" expected_value: 0 - name: Thermal Cycling key: thermal_cycling python: phases.thermal:thermal_cycling depends_on: [vibration_z] timeout: 10h measurements: - name: Thermal Faults validators: - operator: "==" expected_value: 0The duration above is an example value. With on_first_failure left at its default (stop), a unit that fails vibration does not go on to thermal cycling, so each failure is attributed to the first stress that found it. On the dashboard, the failure Pareto then compares Vibration Faults and Thermal Faults across units. If one stress never fails a unit over a significant production volume, while escapes of its defect class stay at zero in the field, that stress is a candidate for removal. For the wider picture of failures that a screen does not catch, see why yield drops.
Common Pitfalls
Choosing a stress by what equipment is available. A chamber on site does not make thermal cycling the right screen for a product whose defects are on heavy wires.
Testing only after the stress. Both stresses produce intermittents that close again at rest and at ambient. Monitor during the stress.
Ignoring the fixture. A vibration fixture resonance inside the test band changes the level at the product more than the profile setting does.
Counting chamber time as product time. The dwell starts when the product reaches temperature, not when the air does.
Transferring a profile between table types. Repetitive-shock and electrodynamic excitations are not equivalent. Redevelop the profile.
Key Points
- Random vibration loads attachments in proportion to mass and resonance: it finds loose hardware, cracked joints on heavy parts and wires, and connector defects.
- Thermal cycling loads bonded joints in proportion to CTE mismatch, temperature swing and package size: it finds solder fatigue, via cracks and component defects.
- Some defects only appear under both stresses together, which combined profiles address.
- Vibration is short per unit but needs a product fixture; thermal cycling is long but loads many units at once.
- Choose from field returns and product construction, then confirm with data by recording each stress as its own phase with its own fault measurement.
