A populated circuit board can pass every test the day it is built and still carry a flaw that will open weeks later, a solder joint barely holding, a hairline crack under a chip. Inline reliability screening catches that by thermal cycling the assembled board, swinging it cold to hot fast enough that the mismatch in expansion between the joint and the parts around it fatigues a weak joint to failure in hours, where the field would take months. The board runs powered and watched through the swing, so an intermittent fault shows in the act. This is environmental stress screening, applied at the board, before the weak one is built into a product.
A board newly assembled and tested has not been proven reliable. The tests it has passed show that it works now, that every part is present, placed, and functioning the day it left the line. They say nothing about the joint that is holding by a thread, the crack too fine to see, the contact that will let go the first time the board runs warm in a customer’s hands. Those are the early failures, the small share of boards carrying a latent flaw that will surface in the first weeks of use, none of which a functional test taken at room temperature finds, because at room temperature the flawed board still works. Reliability screening exists to find them before they ship, by stressing every board hard enough and in the right way to make a weak one fail in the factory, before it can fail in the field. For a populated board, the stress that does this is not steady heat. It is thermal cycling, the repeated swing from cold to hot that works the board the way years of real use would, compressed into hours.
Environmental stress screening, or ESS, is the practice of exposing a newly built electronic assembly to stresses, chiefly thermal cycling and vibration, to force its latent defects to show themselves as failures during the screen. The aim is precise. It is to make the boards that would have failed early fail now, on the line, where they can be pulled, so the boards that ship have already survived the window in which failure is likeliest. Two things screening is not are worth stating plainly, because both are easy to assume. Screening is not a test of the design. A design’s soundness is proven elsewhere, by qualifying it against its requirements; screening takes a sound design as given and looks for the units that were built wrong. Screening also does not make a board more reliable. It cannot improve a board that passes, since it adds nothing and strengthens nothing. What it does is remove the weak boards from the population, so the ones that remain are better as a group, their early failures already culled. The distinction matters because a screen oversold as a reliability booster invites the wrong expectations. A screen finds and removes the weak. The strength of the survivors is the work of the design and the build, decided before the board ever reaches the chamber. Holding this line is part of using a screen well. A maker who believes a screen makes boards reliable may skimp on the design or the assembly, trusting the screen to fix what it cannot. The screen only sorts the boards it is given into the weak and the sound; it cannot turn a poorly built board into a good one. A good screen on top of a good build removes the early failures a sound process still leaves behind. A screen used to paper over a bad process catches more, ships fewer, and never reaches the yield a better build would have given for less.
A steady soak at one high temperature is the right stress for some screens, and the wrong one for a populated board’s commonest failure. The dominant way a board fails over time is not that a part cooks. It is that a solder joint fatigues and cracks. That comes from temperature changing, where steady high heat leaves the joint untouched. The reason lives in a mismatch of expansion. A chip package, the board it sits on, and the silicon die inside the chip each expand by a different amount when heated, set by their different coefficients of thermal expansion. When the assembly warms, the parts grow at different rates, so the solder joints that tie them together are strained by the difference, sheared a little as the materials pull against one another. Cool it down and the strain reverses. Each swing from cold to hot and back works the joint once. A joint worked enough times fatigues, the way a wire bent back and forth enough times breaks. Thermomechanical fatigue of this kind is the single largest cause of board failure, by a wide margin. A weak joint, one already marginal from a poor reflow or a void, fails after only a few cycles, where a sound joint would take many thousands. Thermal cycling is built around that fact. It swings the board cold to hot over and over, deliberately, to make the difference in expansion do in hours what a customer’s daily warming and cooling would do over months and years. A steady soak, however hot, never works the joint this way, because it never changes, and a joint never flexed is never fatigued. The swing is the stress, and the swing is what a steady soak cannot give. It helps to see why so small a difference matters. The amounts parts grow are tiny, a fraction of a percent of their size over a wide swing, and still the solder joint bridging two parts that grow by different fractions has to take up that full difference in its own small body. A large chip makes the strain worse, because the difference in expansion grows with distance from the part’s centre, so the joints at the far corners of a big package move the farthest and are strained the hardest. The same swing that barely troubles a small part can crack the corner joints of a large one, which is why the parts a board screen must watch first are its biggest chips, and the solder balls hidden beneath them. The mechanism is mechanical, a small repeated movement working a joint to failure. It explains why a board screen is built around the motion of temperature where a steady soak offers only a single high mark on the dial.
A solder joint fails from temperature changing, not from temperature being high.
The swing is aimed at the joints and the cracks that thermal strain works first. A solder joint left marginal by assembly, starved of solder, voided, or poorly wetted, carries more strain per swing than a full one, so it fatigues sooner. The joints under large parts feel the strain hardest, since the difference in expansion grows with the distance from the part’s centre, which puts the solder balls at the corners of a big chip among the first to crack. A hairline crack already present grows a little with each swing until it parts. What the screen finds, then, is the population of weak joints and incipient cracks that a sound board does not have, the marks of an assembly that went slightly wrong. The boards that survive the screen are the ones whose joints were made well enough to take the cycling without failing, which is the same population that would have taken years of field use without failing. The screen sorts the two apart by compressing into hours the difference between a good joint and a marginal one, which the field would take months to reveal, the marginal joint cracking on the line where it can be caught. The number of swings a screen runs is chosen to match this. Too few, and a marginal joint that needs more cycles to crack survives the screen and ships, defeating the point. Too many, and the screen begins to use up the life of sound joints as well, eating into the margin of the boards that pass. A screen aims for enough swings to fail the weak with confidence while leaving the sound boards with nearly all their life ahead of them, a balance set from what is known about how the joints fatigue.
A cracked joint does not always stay open. A joint can crack only enough to open when it is cold or hot and close again at room temperature, so a board with such a joint passes a test taken warm and at rest, then opens intermittently in service whenever it reaches the temperature that parts the crack. This is the intermittent fault, and it is among the hardest to catch with a static test, because the board looks sound at the bench. Thermal cycling catches it by taking the board to the temperatures that open the crack, and by watching the board while it is there. A joint that opens only when cold reveals itself when the cycle takes the board cold. A joint that opens only when hot reveals itself at the top of the swing. The screen does not need the joint to fail permanently to find it. It needs only to bring the board to the condition that makes the crack open, with the board powered and monitored, so the momentary fault is seen. A board that passes a static test and still carries an intermittent joint is exactly the unit a thermal-cycling screen is built to catch, the one that looks sound until the temperature finds its flaw.
To catch an intermittent fault, the board has to be alive and observed while it is stressed. A screen that cycled a dead board and tested it only afterward would miss every fault that opens at temperature and closes again, since by the time the board was back at the bench and powered, the crack would have closed and the board would test sound. The board is therefore powered through the cycle, running a load that exercises its functions, and monitored continuously, so a fault that appears for a moment at the cold or hot extreme is seen the moment it appears. The monitoring watches for the signatures of a fault, a signal that drops, a function that fails, a current that spikes, any of which marks the board as having shown a defect during the run. This live watching is what turns thermal cycling from a stress into a screen. The cycle provokes the fault; the monitoring catches it, the two working together so that a board which would only ever misbehave at an extreme of temperature is found at exactly the moment the cycle takes it there.
How fast the temperature changes, the ramp rate, is what sets how hard a thermal cycle works the board. A faster ramp drives the strain into the joints faster, raising the stress each swing applies, so a screen that swings quickly between extremes works the joints harder than one that drifts slowly. The ramp rate, more than the extremes alone, is the dial that sets the screen’s severity. There is a limit to how fast it pays to go. Push the ramp too fast and the stress stops being the steady fatigue the screen aims for and becomes a mechanical shock, a different and harsher mode that can damage sound joints along with weak ones. The art is a ramp fast enough to fatigue the weak joints in a useful time, and short of where it shocks the good ones. The cost of a fast ramp, in the chamber it demands and the energy it burns, is a subject of its own, since reaching a high ramp rate is one of the costliest things a chamber can be asked to do. What matters to the screen itself is that the ramp rate is the chief stressor, the setting that decides whether the cycle only warms the board or genuinely works it.
Two related practices sit behind a production screen, worth telling apart because they are often confused. HALT, highly accelerated life testing, is a design-phase exercise. It applies stresses well beyond the product’s rated limits, ramping temperature and vibration up until the product stops working and then until it breaks, to find the margins between where it is meant to operate and where it gives way. HALT is destructive by intent, run on a few units to learn the limits. Its output is knowledge of how much stress the design can take. HASS, highly accelerated stress screening, is the production screen that knowledge makes possible. It applies stresses inside the limits HALT found, hard enough to precipitate the latent defects in every unit, and short of the levels that would damage a sound one. HALT finds the window; HASS screens within it. A screen set without that knowledge is a guess, either too gentle to catch the weak boards or too fierce to spare the good ones. The pairing is what lets a screen be both effective and safe, stressing every board to the edge of what a good one tolerates, and no further, so the weak fail and the sound come through unharmed. The two are easy to confuse because they use the same kinds of stress, temperature and vibration pushed hard, but their purpose is opposite. HALT seeks the limits and accepts breaking the unit to find them; HASS stays inside those limits and must never break a good unit at all. One is a question asked once of a design, the other a routine run on every board built. Treating a production screen as if it were a limit test, by pushing it past what HALT proved safe, would damage the boards it is meant to protect, which is why the two have to be kept distinct in both intent and setting.
Thermal cycling is the chief stress for a board, and it is rarely the only one. A full screen often adds vibration, shaking the board across a range of frequencies while it cycles, because some defects answer to mechanical motion more than to heat. A connector not fully seated, a component with a cracked lead, a heavy part on a long lead that resonates, a cold joint that a shake will break, these reveal themselves under vibration in a way thermal cycling alone might miss. Combining the two, heat and shake at once, is the heart of a thorough stress screen, since a board has to survive both the thermal working of its joints and the mechanical working of its mountings in real use. The chamber that screens boards is therefore often a chamber that can both cycle temperature and vibrate the load, applying the two stresses together so a defect that answers to either is brought out. Vibration and thermal cycling are the two pillars of environmental stress screening. A board put through both has been worked in both the ways that field life will work it. The vibration a screen uses is usually random, its energy spread across a band of frequencies at once, so that whatever frequency a given board is sensitive to is excited without having to be found first. A single swept tone would have to pass through each board’s weak frequency in turn; broadband shaking reaches them all together, which suits a screen meant to work on every board the same way.
A bare board is awkward to handle in a way a finished product is not. It has no case to grip, its surface is crowded with parts that must not be touched, and it carries a static charge risk that a careless touch can discharge into a component and damage it. Fixturing a board for the screen has to solve all of this at once. The fixture holds the board without flexing it, since a long board left to sag under its own weight is stressed in a way the screen did not intend, and supports it so the vibration shakes the entire board, where an unsupported edge would only bounce. It protects against static discharge, grounding the handling so a charge cannot build and jump. It brings power and monitoring to the board through edge connectors or a bed of probes, so the board can run and be watched while it cycles. The fixture also sets how many boards a chamber screens at once, racking them so the air reaches each evenly and the robot can load and unload them. Designing the fixture is much of the work of bringing a board screen to a line, because it is the part that touches the bare board and decides whether it can be stressed, powered, and watched without being harmed in the handling. The fixture also has to take the same cycling and vibration the board does, run after run, without loosening or failing itself, since a fixture that cracks or rattles loose becomes a source of false failures, marking sound boards as faulty when the fixture has lost a connection that the board has not. A fixture for a stress screen is therefore built to outlast the boards it holds, rugged enough that it is never the weak link in the screen it serves.
Screening at the board level is cheaper than catching the same fault later, which is much of the reason to do it inline. A weak joint found on a bare board is repaired by reworking the joint, a small fix on an accessible board. The same joint, left to pass and built into a finished product, fails in the field after the board is buried inside an enclosure, wired to other parts, and shipped. Finding it then means a return, a teardown to reach the board, and often the scrap of an entire product around one bad joint. The cost of a fault climbs steeply at each step it survives, from a rework on the line, to a teardown in the factory, to a return from a customer. Screening the board before it is built into anything catches the fault at the cheapest point, while the board is still bare and the fix is still small. The finished product has its own burn-in, which screens the assembled product for the faults that only the assembly introduces. The board screen catches the board’s own faults first, before they are sealed inside something larger and costlier to open.
What a board screen delivers, in the end, is a population with its weak members removed. Every board that ships has been swung cold to hot, powered and watched, hard enough to crack a marginal joint, and has come through with its joints intact. The boards that did not come through are pulled, their flaws found on the line, before they could reach a customer’s hands. This is the sum of what the screen does, and it is worth being clear that it is no more than this. The screen does not strengthen the boards that pass; their strength was built into them. It does not test the design; that was proven before production. It removes the early failures from the shipped population, and tracks which board saw which screen, so each board’s history is on record alongside the rest of its build. A board that comes through a stress screen is not a better board than it was. It is a board now known to be free of the early failures the screen provokes, which is exactly what a reliability screen is for, and exactly the boundary of what it can promise.
Because the commonest way a board fails over time is a solder joint cracking from fatigue, which comes from temperature changing, where a steady high temperature never works the joint at all. The chip package, the board, and the die expand by different amounts when heated, so each swing from cold to hot strains the joints that tie them together. A weak joint fatigues and cracks after a few swings, where a sound one survives thousands. A steady soak never flexes the joint, so it never finds this fault.
HALT, highly accelerated life testing, is a design-phase test that stresses a few units beyond their rated limits to find where they give way, learning the product’s margins. It is destructive by intent. HASS, highly accelerated stress screening, is the production screen built on that knowledge, stressing every unit inside the limits HALT found, hard enough to precipitate latent defects, short of harming a sound board. HALT finds the window; HASS screens within it.
By powering the board and monitoring it continuously while it cycles. An intermittent joint may open only at a temperature extreme and close again at room temperature, so a static test at the bench misses it. Taking the board to the cold or hot extreme that opens the crack, with the board running and watched, makes the momentary fault show at the instant the cycle reaches the temperature that parts the joint.
No. Screening removes the weak boards from the population; it does not improve the boards that pass. Their reliability was built into them by the design and the assembly. A screened board is not stronger than it was, only known to be free of the early failures the screen provokes. A screen oversold as a reliability booster invites the wrong expectations; its real job is to find and pull the weak before they ship.
Because catching a fault on a bare board is far cheaper than catching it later. A weak joint found on the board is reworked easily; the same joint left to fail inside a finished product means a return, a teardown, and often a scrapped product around one bad joint. The cost of a fault climbs at every step it survives, so the board screen catches the board’s own faults at the cheapest point, before they are sealed inside something larger.
Envsin thermal-cycling and vibration chambers for inline PCBA stress screening, reliability testing and production integration.