HALT produces two numbers. The Fundamental Operating Limit (FOL) is the stress level at which the product stops functioning. The Destruct Operating Limit (DOL) is the stress level at which the product is permanently damaged. Everything that happens in a HALT chamber — every thermal step stress, every vibration step stress, every combined stress level — is aimed at finding these two numbers for temperature, vibration, and their combination.
Most teams find them, write them down in the HALT test report, and move on to qualification. That is the wrong thing to do with the data. Here is what to do instead.
What FOL and DOL actually mean
The FOL is not a specification limit. It is an observation about the current design. If the FOL for cold temperature is -62°C and the specification requires operation to -40°C, the margin is 22°C. That margin is the current design's headroom between what it must do and what it can do. It says nothing about whether the design is good or bad — only about how much room exists between the specification and the failure point.
The DOL is the point of no return. Below the FOL and above the DOL, the product is non-functional but returns to specification when stress is removed. Below the DOL, the product does not recover. The distance between FOL and DOL — the stress margin — tells you something about the robustness of the failure mechanism at the FOL. A FOL at -62°C and a DOL at -65°C means there are only 3°C between non-function and permanent damage. A FOL at -62°C and a DOL at -78°C means there is 16°C of additional headroom. Same FOL, very different behaviour at the boundary.
The six things to do with HALT data
1. Calculate the margin to specification — and act on it if it is narrow. The rule of thumb used by most experienced HALT practitioners: the FOL should be at least twice the specification limit from ambient. If the spec requires -40°C operation and the FOL is at -62°C, the margin is 22°C. If the ambient is 25°C, the spec is 65°C below ambient, and the FOL is 87°C below ambient — a ratio of 1.34. Less than 2.0 is a yellow flag. Less than 1.0 means the product cannot reliably meet its specification under production variation. These numbers need to be calculated per axis (cold, hot, vibration Grms) and the narrowest margin is the design's limiting constraint.
2. Identify the failure mode at the FOL — not just the symptom. If the product fails electrically at -62°C, the failure mode is not "cold temperature failure" — it is something specific: a crystal oscillator that stops oscillating, a capacitor whose value shifts enough to pull a voltage regulator out of regulation, a mechanical relay that sticks. Find the physical mechanism. The corrective action for a crystal oscillator failure is different from the corrective action for a capacitor shift. A HALT test that identifies failure symptoms without identifying failure mechanisms produces data you cannot act on. The failure mode analysis connects directly to the field failure correlation framework.
3. Decide whether to fix or accept each failure mode. Not every failure mode found in HALT requires a design change. If the FOL is well outside the specification limit and the failure mechanism is benign — it recovers completely when stress is removed and the margin to specification is large — the correct decision may be to accept the failure mode and move on. HALT is a discovery tool, not a compliance test. You are not required to eliminate every failure mode it finds. You are required to make a conscious, documented decision about each one: fix it, monitor it, or accept it with rationale.
4. Use the FOL and DOL to set HASS screen conditions. HASS (Highly Accelerated Stress Screening) is the production application of HALT data. The HASS screen applies stresses between the FOL and the DOL — high enough to precipitate latent defects introduced during production, low enough not to consume life from good product. The HASS stress levels are derived from the HALT FOL and DOL: typically 50–80% of the FOL margin from the specification limit. Without HALT data, HASS screen conditions are guesses. With HALT data, they are derived from demonstrated product capability. The HASS methodology is covered in HASS Testing: The Test That Catches What Your Production Line Misses.
5. Track FOL and DOL across design iterations. HALT is most valuable when it is run repeatedly — at each major design revision, each supplier change, each PCB layout change. The FOL and DOL from the previous HALT are the baseline. If the new design's cold FOL moves from -62°C to -55°C, the margin to specification has narrowed. That is a signal worth investigating before the product ships — not after it fails in the field. HALT data accumulated across design generations is a design history that tells you which changes helped and which hurt.
6. Document the test conditions, not just the results. HALT results are only reproducible if the test conditions are documented with enough precision to repeat them. The thermal step stress profile (step size, dwell duration, criterion for defining equilibrium), the vibration step stress profile (Grms level, step size, dwell duration), the combined stress sequence, the DUT configuration (powered, monitored, functional test interval) — all of this goes in the test report. A HALT report that says "product failed at -62°C" without documenting how it got to -62°C cannot be used to design a repeat test, cannot be used to set HASS conditions, and cannot be used to verify that a redesign actually improved the margin. The documentation requirements for HALT records are part of the broader test record requirements.
The question nobody asks but everybody should
What happens to the margin when the product ages? The FOL and DOL are measured on new, unaged hardware. After 5 years in the field — after some number of thermal cycles, some humidity exposure, some mechanical fatigue — the margin will be different. In most cases, it will be narrower: capacitors shift, solder joints crack partially, connectors develop contact resistance. The FOL measured in HALT is the initial margin. The design must have enough initial margin that the aged margin — after the expected field life — still exceeds the specification limit.
This calculation requires either prior data on how aging affects the FOL for similar products, or a physics-based model of the failure mechanism's progression over time. It is rarely done rigorously. When field failures occur in aging products that passed qualification, this is frequently the gap that was not closed.
Reading a HALT chart the way an experienced practitioner does
A HALT step-stress chart shows the product's response at each stress level, typically a plot of step number against a measured parameter with markers for the FOL and DOL. Most engineers new to HALT read this chart left to right, watching for the first failure marker. Experienced practitioners read it for shape, not just the failure point. A product that performs flawlessly through every step until a sudden, catastrophic failure at the FOL is telling a different story than one that shows progressive parametric drift for several steps before failing — the first suggests a discrete failure mechanism with little warning margin, the second suggests a mechanism with observable precursors that a production test could potentially screen for even before HASS is designed. The shape of the approach to failure is often more actionable than the failure point itself.
The HALT result that gets thrown away by mistake
When a HALT test ends with multiple failure modes found across different stress axes, it is common practice to report only the limiting one — the lowest FOL across temperature, vibration, and combined stress — because that is the number that defines the product's overall margin. This is correct for setting an overall margin figure, and it discards information that matters for a different purpose: design improvement. A product with a cold temperature FOL of -55°C and a vibration FOL well beyond any field requirement has a clear, single area to improve. A product with FOLs clustered close together across multiple axes is differently fragile — improving any one axis without addressing the others leaves the product limited by whichever axis comes next. Keep the full multi-axis result, not just the worst number, in the design record. The single limiting figure goes in the summary; the full picture belongs in the file an engineer opens before starting the next design revision.
Comparing HALT results across supplier-sourced components
When a design change involves switching a component supplier — a different connector, a different crystal oscillator, a different voltage regulator — re-running HALT on the new configuration and comparing the resulting FOL directly against the previous supplier's result is one of the few ways to catch a supplier-driven reliability regression before it reaches volume production. A 12°C narrower margin on the new supplier's component, found in a half-day HALT comparison, is a far cheaper discovery than the same gap found through field returns eighteen months after a supplier change that looked, on every datasheet, like a drop-in replacement.