Robotic Chamber For IC Final Test Handler Integration

Prove it across the range

A datasheet's promise has to be proven.

What final test is

Final test is the last gate a chip passes before it ships. After the die is packaged, every unit is handed to automated test equipment that runs it against its datasheet, checking that it works, measuring how well, and deciding which grade it earns.

The handler is the machine that serves that test, taking packaged parts from their trays, bringing them to the tester one batch at a time, and sorting each by the verdict it earns into the bin its result calls for, the good graded, the bad categorised, the marginal set aside.

Why test at temperature, and why it is hard

The reason a final-test handler carries a thermal system at all is that a chip's datasheet makes a promise across temperature, and a promise has to be proven. A part rated to work from minus forty to a hundred and twenty-five degrees cannot be trusted on the word of a room-temperature test alone, since the things that drift with heat, the timing, the leakage, the exact voltage at which a gate switches, are the things a customer will lean on at the edges of that range. Final test answers that by checking each device hot and cold as well as warm, and the hard part is not the testing but the holding: the device has to be at the target temperature at the instant the tester contacts it, and stay there for the brief moment the test takes. Getting it there is done one of two ways, and the choice shapes the whole machine. The older way is to soak the parts, holding them in a heated or chilled boat until their mass reaches the setpoint, then presenting them to the contactor and trusting them to stay put through a test measured in milliseconds. That trust is the weakness: a low-power part holds, but a device that dissipates real power heats itself the instant it is energised, climbing off the cold setpoint or overshooting the hot one before the test even finishes, so the reading is taken at a temperature the part was never meant to report. The answer to that is active thermal control, a temperature-driven head that clamps onto the device during test and forces it to a target, pumping heat in to hold a cold part cold against the chill, or dragging heat out to keep a hot, self-heating part from running away, so the junction sits where the datasheet says no matter what the silicon does. Cold brings a second enemy, water: test a part below the dew point and the moisture in the air freezes onto the device, the socket, and the handler, and ice on a contact is a false failure waiting to happen, so the cold end of the machine is flooded with dry air or nitrogen to keep the damp out. And around all of it runs the pressure of throughput, because the tester the handler feeds cost more than anything else in the room, and a handler that cannot place a part at temperature, on contact, fast enough to keep that tester busy is burning money with every idle second.

Gravity, pick-and-place, turret

Handlers come in families shaped by the package they move. The gravity handler, the old workhorse, lets parts slide down rails to the test site under their own weight, well suited to the leaded packages that came in tubes.

The pick-and-place handler, the modern standard, lifts parts from a tray with a vacuum nozzle and sets them precisely onto the contactor, the only practical way to handle the leadless and ball-grid packages a gravity rail cannot guide.

For tiny, high-volume parts a turret handler spins them past stations on a rotating table at great speed, and for parts still in strip form a strip handler tests them before they are singulated, each a different answer to the same question of how to move a part to the tester and away.

Warm alone is only half a test

A chip checked only at room temperature is a chip checked only halfway.

Soak versus active thermal control

The two ways to bring a chip to temperature split handlers into two camps. The simpler soaks the parts in a hot or cold boat until their mass settles at the setpoint, then feeds them to the contactor, cheap and fine for a part that draws little power and holds its temperature through the short test.

The trouble is the part that heats itself. A device dissipating watts warms the instant it powers up, so a part soaked to minus forty can be well above it by the time the test ends, and the tester records a pass at a temperature the device never truly sat at.

Active thermal control answers that with a head that grips the device and holds it to a target throughout the test, sensing its temperature and pumping heat in or pulling it out fast enough to cancel the device's own swing. For a high-power chip it is the one honest way to test cold or hot.

The cost is speed and money: a thermal head is slower to engage and dearer to build than a passive soak, so a line tests with soak where it can and reaches for active control where the device's self-heating would otherwise pull it off its setpoint.

The condensation problem at cold

Testing a chip cold opens a door to water. Drop a device and its socket below the dew point of the room and the moisture in the air condenses and freezes on them, and frost on a contact or in a socket reads as a fault the chip never had.

The cold end of a handler fights it by keeping the damp out, flooding the test site with dry air or nitrogen and sealing the cold zone so room air cannot creep in, since a single bead of ice between a pin and a pad can fail a good part, and a dry-purge that lapses can stop a line.

Keeping the tester busy

The economics of final test are ruled by one fact: the tester is the costliest thing on the floor, and it earns nothing while it waits. Everything the handler does is bent toward never letting that happen.

The first lever is parallelism. Rather than test one device at a time, a handler presents many at once, dozens or hundreds in step for a memory part, so a single pass of the tester grades a whole tray of them at once and the tester's costly seconds are spread across many units.

The second is index time, the dead moment between one set of devices leaving the contactor and the next arriving. A handler is engineered to shrink that gap, since every second the tester spends waiting for parts is a second the line is paying for and getting nothing back.

Temperature complicates both. A part has to reach its setpoint before it can be tested, and a handler that stalls the tester while a tray soaks has traded one idle cost for another, so the conditioning is pipelined, the next parts warming or chilling while the current ones are under test.

The number that captures it all is utilization, the share of the time the tester is truly testing, and a handler is judged by how high it can hold that share far more than by its own raw speed.

The contactor, where chip meets tester

Between the handler and the tester sits the contactor, the small, much-abused part that touches the chip. Its spring pins or elastomer pads press onto the device's leads or balls to carry the test signals, and they have to land in exactly the right place, flat and clean, on a package that may bear hundreds of contacts finer than a hair. The contactor wears with every insertion, and a pin gone soft or a pad gone dirty quietly turns good chips into false rejects, so it is watched, cleaned, and replaced on a schedule like the consumable it is.

A bad contact is a false reject

A worn contactor will fail a perfectly good chip, and that false reject costs the line as much as a true one does.

Binning and the device's record

The end of the test is a sort. Each device is dropped into a bin by what the tester found, the passing parts split by grade, the fast silicon to one tray and the slower to another, and the failures split by the kind of fault they showed.

The result follows the part. Tied to the device by its position or its mark, the bin it landed in and the data behind it are logged against that unit and its lot, so a wafer that yields badly or a tester drifting out of calibration shows up in the numbers rather than in a customer's complaint.

Tri-temp, in one pass or three

A part rated across a range has to meet its spec at each end and in the middle, and there are two ways to get there. A handler can take each device through hot, cold, and ambient in one pass before it moves on, or the whole lot can be run at one temperature and then run again at the next, retesting from the start each time. The first spares the parts a second handling; the second is simpler to set up, and which a line chooses turns on its volume, its mix, and how dearly it values the tester's time.

When the handler jams

A handler is a machine of fine motions at speed, and now and then a part lands wrong, a nozzle misses, a tray sticks, and the flow stops. When it does, the tester stops with it, the expensive asset idled by a jam in the cheaper machine that feeds it.

A good handler is built to clear the common jams on its own and to call for help on the rest before the idle minutes mount, since on a line built around a costly tester the handler's value is measured by how rarely it makes that tester wait.

Not a screen, a sort

It helps to be clear about what final test is not. It does not age the chip or try to break the weak ones the way a burn-in or a stress screen does; it asks a different question, whether this part, as built, meets its spec across temperature, and what grade it deserves.

A device can pass final test and still hold a latent flaw that only burn-in would catch, and it can survive burn-in and still fail final test on a parameter out of spec, so the two sit side by side in a flow, one proving the part is good now, the other weeding the ones that will not stay good.

Chip, at temperature, on contact

An IC final-test handler is a thermal machine and a robot fused into one job: to put each chip at the temperature its datasheet is written around, hold it there against its own self-heating, present it to the tester clean and dry and exactly aligned, and sort what comes back, thousands of times an hour without letting the tester idle.

The chamber, here, is not where a part is aged but where it is asked, at every temperature it claims, to prove it is what the label says.