English
Views: 0 Author: Vinci Zhang Publish Time: 2026-11-02 Origin: Site
When a board leaves the reflow oven it has been inspected for how it looks, but not yet for how it behaves electrically. The electrical test is the gate that confirms the circuit actually works — that nets are connected, components are present and valued correctly, and nothing is shorted or open. For high-mix assembly, where many different products run in small batches, the choice between flying probe and in-circuit test is one of the most consequential and most misunderstood decisions on the line. This article compares the two on the dimensions that actually decide the trade: fixtures, test points, program development, throughput, and volume, and ends with a framework you can apply to your own mix instead of copying someone else's answer.
High-mix lines change products constantly, and every change forces a decision about how the new board will be verified electrically. Pick the wrong method and you either pay for fixtures you rarely use or wait days for a test program you needed yesterday. The choice is not "which is better" but "which fits the mix," and for a plant running hundreds of part numbers that distinction is the difference between a test area that pays for itself and one that quietly drains margin.
Optical inspection confirms appearance; electrical test confirms behavior. A board can look perfect and still have an open net, a wrong-value resistor, a reversed diode, or a tombstoned capacitor that optics cannot see. The electrical gate is where those escape the line, and for high-mix work it is the verification that varies most in cost from one product to the next. Treat it as a quality step and a cost center at the same time, because the method you select sets both the escape risk and the per-board burden.
On a single high-volume product, the fixed cost of a fixture spreads across millions of boards and becomes negligible. On a high-mix line, that same fixture cost is amortized across a few hundred boards before the product changes, which flips the calculation. The method that wins at volume can lose badly at mix, and the reason is purely the denominator: when the run length is small, every fixed cost lands hard on each unit. This is why shops that were happy with ICT on a former high-volume program often find it unworkable the moment they move to contract or prototype work.
The electrical test method affects test-point requirements, programming effort, and line balancing, all of which connect to the inspection plan you built from solder paste to final test. Treating the choice in isolation produces a test station that does not fit the line around it. A flying probe that is too slow becomes a buffer-eating bottleneck; an ICT that demands test points the design team never reserved becomes a coverage gap. The test method is a line-design input, not a late add-on.
We will compare flying probe and ICT on fixtures, test points, program development, throughput, and cost, and end with a framework for choosing. We will not recommend one universally, because the answer depends on your mix, your volumes, and how often your products change, which is exactly the point. Where a second method such as first article inspection at changeover with the FAI1095 machine, or data-driven defect feedback, complements the electrical gate, we will say so. The broader question of which inspection methods to deploy, and how to match them to defect risk, is covered separately in how to choose SMT inspection equipment and matching methods to defect risks; here we stay focused on the electrical test decision, because the electrical test is one layer in a larger verification system rather than the whole of it.
A flying probe tester verifies a board with a small number of moving probe heads that travel to test points under program control. There is no fixture built for the board; the "fixture" is the software and the probes' motion path. That single fact drives every strength and weakness that follows, and it is the reason flying probe is the default answer for shops that change products more often than they repeat them.
Because nothing physical is built for the board, there is no bed-of-nails to design, fabricate, or store. The tester contacts pads, vias, and component leads directly under program control, which means a new product can be tested as soon as its program exists, without waiting on a fixture shop. The absence of hardware also removes an entire class of failure: a worn pin, a mislocated probe, or a fixture that no longer matches a revised board. For a high-mix shop, that freedom from physical tooling is often worth more than any speed advantage the alternative offers.
Programs are generated from the CAD netlist and can be revised quickly when the board changes. For high-mix shops where designs are in constant flux, this agility is the main reason flying probe is chosen: the test keeps pace with the product instead of lagging it. A revision that would require a new fixture on an ICT line becomes a netlist reload and a re-optimization on a flying probe, measured in hours rather than weeks. When your design churn exceeds your production volume, that responsiveness is the deciding variable.
Without a fixture to pay for, the marginal cost of adding another product to the test roster is small. That makes flying probe the natural fit when the number of distinct products is large and the volume of each is modest, which is the textbook high-mix profile. The accounting is simple: each new part number costs a program, not a fixture, and a program is mostly engineering time rather than physical inventory. Shops with catalogs in the hundreds or thousands of part numbers rarely want that many fixtures sitting on a shelf.
The trade-off is speed. Moving probes to hundreds or thousands of points takes time, so a flying probe pass is slower than a parallel ICT hit. At low volume that slowness is acceptable; at high volume it becomes the bottleneck that argues for a fixture. A board with several thousand test points may take minutes rather than seconds, and on a fast SMT line that gap forces buffering or a second tester. The method is not slow in absolute terms — it is slow relative to a contact-everything-at-once fixture, and that relativity is what sets the crossover volume.
In-circuit test presses the board onto a bed-of-nails fixture, a plate carrying a spring-loaded pin for every test point, and measures the whole board in essentially one parallel contact. The fixture is the defining asset and the defining cost, and understanding that duality is the key to using ICT well instead of resenting it.
A board on a well-made fixture is tested in seconds, and the test can cover analog, digital, and powered conditions thoroughly. For a product built in volume, ICT delivers a depth and speed that flying probe cannot, which is why volume lines standardize on it. The parallel nature of the contact means adding more test points does not linearly add test time the way it does for a moving probe, so complex boards benefit most. When the same product runs for months, the fixture cost evaporates into the volume and the speed advantage compounds into real savings.
ICT can power the board and run functional sub-tests at the point of contact, catching defects that require the circuit to be live. Flying probe can do some of this, but the parallel, powered nature of ICT makes it the deeper test where fixture economics allow. Power-on testing reveals a class of failure — a regulator that does not regulate, a oscillator that does not start, a bias network that is wrong — that a passive continuity check alone will miss, and that depth is precisely why automotive and medical programs often mandate ICT.
Because every point is contacted at once, ICT is fast and repeatable, and it can apply power and exercise circuits in ways a flying probe cannot easily match. The same fixture, however, must be designed, fabricated, debugged, and stored for each product, which is the cost that high-mix lines feel most. A single fixture can cost as much as a meaningful fraction of the tester and can take weeks to mature, so the burden is both financial and temporal. The fixture is what makes ICT fast and what makes it expensive, and you cannot have one without the other.
The fixture cost is largely independent of how many boards you build, so its per-board burden falls as volume rises. Below a certain volume it is simply a sunk cost that flying probe avoids, and that threshold is the heart of the decision. Estimating that threshold is not guesswork: divide the fully loaded fixture cost by the expected board count, add the program and debug time, and compare to the flying probe program cost plus its slower per-board time multiplied by volume. The method with the lower total at your actual volume wins.
Almost every flying-probe-versus-ICT argument reduces to one question: how many boards will this fixture test before the product changes? The fixture is where the two methods diverge most sharply, and it is where the money is decided. Everything else — speed, coverage, convenience — is secondary to this single economic pivot.
If a product will run a few hundred boards before changing, the fixture cost per board is high and flying probe wins. If it will run tens of thousands, the same fixture cost per board is tiny and ICT wins. The crossover is not a fixed number; it moves with fixture price and your labor rates, but the direction is always the same. A useful rule of thumb many shops use: if expected lifetime volume is under roughly one to two thousand boards, flying probe is usually cheaper; above ten thousand, ICT usually is; between them, decide on lead time and flexibility as much as on unit cost.
A flying probe program can exist within a day of receiving the netlist; an ICT fixture may take weeks to design and debug. In a high-mix environment where products appear faster than fixtures can be built, that lead-time gap alone settles many decisions, regardless of unit cost. A customer who needs a hundred boards next week cannot wait a month for a fixture, and the speed-to-first-test of flying probe is then worth more than any per-board saving. Lead time is a cost too, paid in expedite fees and lost opportunity rather than on an invoice.
Every ICT fixture is a physical object that must be stored, maintained, and eventually repaired or retired. High-mix shops accumulate fixtures faster than they use them, and the storage and upkeep become a hidden cost that flying probe simply does not have. A fixture for a product run once a year still occupies space, still drifts out of calibration, and still needs a technician who remembers how it was built. Multiply that by a few hundred part numbers and the real estate and labor alone can exceed the tester's depreciation.
Some lines keep both: flying probe for the long tail of low-volume products and ICT for the few high runners. The plan for inspection locations on the line should reflect that split, placing each method where its economics work rather than forcing one onto every product. The acceptance references most factories work to, such as those published by IPC, give both methods a shared definition of test coverage that survives a customer audit, so a hybrid strategy does not confuse the conversation about what "tested" means.
Both methods need access to the circuit, but they need it differently, and that difference reaches back into how the board is designed. The test method should be chosen before layout, not after, because retrofitting access is expensive. A decision made at schematic capture about test access can save a fixture redesign later.
Because probes move, flying probe can reach pads, vias, and leads that a fixed fixture pin cannot physically occupy. This forgiving access is a real advantage on dense or late-changing designs where reserved test points were not planned. It also tolerates boards where the original test strategy was an afterthought, which is common on legacy or customer-supplied designs. The flexibility does not remove the need for access entirely, but it widens the set of boards that can be tested without a redesign.
A bed-of-nails needs a clean, unobstructed pad under every planned pin, which means the board must reserve test points during layout. Skipping that discipline forces compromises in the fixture and reduces coverage exactly where it matters. Test-point placement, keep-out zones around tall components, and avoiding probes landing on heatsinks or connectors are layout rules that must be enforced early. Shops that run ICT successfully treat test-point insertion as a design-rule check, not a favor the test engineer asks for after the fact.
There is a permanent tension between packing the board tightly and leaving room to test it. ICT makes that tension explicit and early; flying probe relaxes it but cannot fully escape it, because probes still need somewhere to land. Dense RF and miniaturized designs push both methods, and at some pitch only in-circuit access via test pads or boundary scan remains viable. Recognizing this tension before layout lets you choose a test method whose access demands the design team can actually meet.
Shops that design once and build forever can plan test points rigorously and favor ICT; shops that design constantly and build a little favor flying probe. The test method is as much a match to the design culture as to the product. If your engineers will not reliably reserve test points, ICT will underperform no matter how good the fixture house is; if your products change weekly, flying probe will keep up where ICT cannot. Match the method to how your organization actually behaves, not to how you wish it did.
The program is the other dividing line, and it behaves opposite to the fixture: flying probe is cheap to start and easy to change, ICT is expensive to start and costly to change. Where the fixture decides the volume economics, the program decides the responsiveness economics, and high-mix shops live or die on responsiveness.
Generated from the netlist and tuned visually, a flying probe program can be running the same day the data arrives. For high-mix shops that is the difference between testing a new product this week and next month. The visual tuning step — confirming probe landing, adjusting for tall components — is engineering time, not tooling time, so it scales with staff rather than with capital. That is exactly the kind of cost a variable-volume operation wants.
ICT program development includes fixture debugging, which means the first good test may arrive weeks after the board does. Once there, however, the program is fast and deep, which is the payoff that volume lines wait for. The richness comes from powered tests, analog measurements, and functional sub-routines that are practical only when the board is held firmly by a fixture. The wait is the price of that depth, and on a stable high-runner the price is worth paying.
When a board revises, a flying probe program updates with the netlist; an ICT fixture may need rework or replacement. In a world of constant revisions, that flexibility is often the deciding factor more than raw unit cost. A board that goes through several spins in a year can obsolete a fixture mid-life, whereas a flying probe simply ingests the new netlist. For prototype and new-product-introduction work, this single property frequently settles the argument.
ICT programs and fixtures are tied to a specific board revision, and the discipline to retire or update them as products change is itself a cost. Flying probe trades that discipline for a per-board speed penalty, which is the cleaner trade for many high-mix shops. Either way, the test results should feed the same defect data and reporting loop as the rest of the line, because the value of electrical test is not only the pass or fail but the trend it contributes to yield improvement over time.
Strip away the detail and the choice is a curve: flying probe is cheaper at low volume and slow at high volume; ICT is expensive to start and fast forever after. The job is to find where your products sit on that curve, and to remember that "volume" means volume per fixture, not volume of the building.
Below a few hundred boards per product, the absence of fixture cost makes flying probe the lower total cost almost by definition, and its slower per-board speed is irrelevant when the run is short. This is the high-mix sweet spot, and it is also the prototype and new-product zone where speed-to-first-article matters more than cost per unit. For these runs, paying for a fixture would be paying for capacity you will never use.
Above tens of thousands of boards, the fixture cost vanishes into the volume and the speed advantage of parallel test compounds into large savings. This is the high-volume sweet spot, and forcing flying probe here is leaving money on the table. The per-board time difference, multiplied across a long run, can mean an extra tester or a missed ship date, so the fixture investment pays for itself quickly and then keeps paying.
The difficult zone is the middle, where neither advantage is overwhelming, and the decision should be made on lead time and flexibility rather than unit cost, because the cost gap is small and the agility gap is large. Around one to ten thousand boards, a careful total-cost model is mandatory, and the answer often hinges on how reliably you can forecast the run length. If the forecast is soft, the flexibility of flying probe is the safer bet even if the unit math slightly favors ICT.
A slow flying probe station can unbalance a fast SMT line, forcing buffers or a second tester. The inline versus offline decision for the test station affects whether that bottleneck stalls the whole line or sits quietly beside it. An ICT in the line is fast enough to keep pace; a flying probe may need to be offline, or doubled, to avoid starving downstream. Balancing the test station against the line cycle time is part of the same calculation, not a separate one.
The choice is not a preference but a calculation across five inputs: volume per product, number of products, revision frequency, test-point discipline, and how the station fits the line. Answer those and the method usually selects itself, and you can defend the choice to a customer or an auditor with numbers rather than opinion.
Estimate boards per product over its life. Low and infrequent points to flying probe; high and steady points to ICT; the middle points to whichever constraint — lead time or flexibility — hurts more. Be honest about the forecast: a "high volume" program that slips to a trickle still owes its fixture cost, while a "low volume" program that ramps can outgrow flying probe. Revisit the assumption quarterly as the mix shifts.
A large, growing product catalog argues for flying probe's low marginal cost; a small, stable catalog argues for ICT's depth. The product count is the multiplier on every fixture decision you make. A shop with fifty part numbers and one big runner may still justify ICT for the runner while flying-probing the rest; a shop with two thousand part numbers almost never fixtures them all. The catalog shape, not the headline volume, drives the answer.
Frequent revisions punish ICT's fixture rework and reward flying probe's netlist-driven updates. If your designs change more than your volumes grow, flying probe's flexibility usually wins the argument. Track how many engineering changes per product per year you typically see; above a threshold, fixture churn alone can erase ICT's unit-cost advantage. For new-product-introduction and contract work, this single metric is often decisive.
If your boards are designed with reserved, accessible test points, ICT is viable; if they are not, flying probe will serve you better until the discipline exists. Choose the method your design culture can actually support. If you want ICT but your layouts rarely reserve test points, the honest move is to fix the design process first, or to accept reduced coverage — do not buy the fixture and hope the access appears.
Once chosen, the FAT and SAT checklist for inspection equipment still applies: confirm coverage, speed, and data export at acceptance, and confirm the program actually exercises the defects you care about before the line depends on it. A test method chosen well and accepted poorly still fails the board. Verify the program against known-good and known-bad samples, confirm the data it emits is usable, and only then let the line treat a green light as real.
What is the main difference between flying probe and ICT? Flying probe tests a board with moving probes and no fixture, so it is fast to set up and flexible but slower per board. ICT uses a bed-of-nails fixture that contacts every point at once, so it is fast and deep but requires a fixture to be built and stored for each product.
Which is better for high-mix, low-volume assembly? Flying probe is usually better when each product runs in small batches, because it avoids fixture cost and lead time. ICT only pays off when a single product runs in high enough volume to amortize the fixture.
Does flying probe need test points like ICT does? It is far more forgiving: flying probes can reach pads, vias, and leads that a fixed fixture pin cannot occupy. ICT requires disciplined, reserved, unobstructed test points planned during board layout.
How long does it take to program each method? A flying probe program can often be generated from the netlist and running within a day. An ICT program includes fixture design and debugging and may take weeks before the first good test.
Can ICT test powered and functional conditions? Yes. Because ICT contacts the whole board at once and can apply power, it can run functional sub-tests that catch defects requiring the circuit to be live, which is one of its main advantages over flying probe.
Should we keep both methods on one line? For many high-mix operations a hybrid works well: flying probe for the long tail of low-volume products and ICT for the few high runners. The key is placing each where its economics work rather than forcing one method onto every product.
