Publish Time: 2026-10-28 Origin: Site
The inline versus offline decision changes how a line behaves every single shift, and it is usually made for the wrong reason: whichever configuration a supplier happened to quote. Inline inspection sits inside the conveyor and sees every board automatically; offline inspection sits beside the line and depends on somebody carrying boards to it. Both are legitimate, most real factories run a mix of the two, and the right mix follows from throughput, coverage requirements, handling constraints, labour, and budget rather than from habit. This article lays out the decision in the order that actually matters and gives you a framework for assigning each inspection checkpoint to an inline or an offline SMT inspection equipment configuration.
The distinction is not about capability. The same detection principle can be delivered either way, and the choice is about how the board reaches the machine and what happens to the result afterwards.
An inline machine is built into the production flow. Boards enter and leave on the same conveyor, communication with upstream and downstream equipment is handled automatically, and inspection happens at line rate without an operator. The machine becomes one more station in the sequence, and the line cannot run at full rate unless the inspection station keeps up.
An offline machine stands beside the line. An operator loads boards, starts a program, and reviews the results. It is independent of line rate, so cycle time does not constrain it, but throughput is limited by how many boards a person can carry, load, and unload. Offline equipment is commonly used for sampled audit, failure analysis, new product introduction, and repair verification.
Because inline inspection runs unattended, its results must be consumable automatically: boards routed to a pass or fail destination, alarms surfaced to a responsible person, and data written to a shared system. Offline inspection produces results that a person is already looking at, so the workflow is inherently manual. Choosing a configuration therefore means choosing a workflow, not just a cabinet.
Most lines are not uniform. A high-volume product running for weeks benefits from inline coverage at every checkpoint, while a short-run or prototype job does not justify tying up line capacity for inspection programming. A mixed configuration lets each checkpoint be placed where it costs least, which is the practical answer for most operations rather than a single global choice.
Configuration is decided after the checkpoints are decided. First establish which failures need to be caught and where each becomes visible, as described in our article on building an SMT inspection plan from solder paste to final test, then decide per checkpoint whether it runs inline or offline. Reversing that order produces equipment that is well configured for the wrong job.
Cycle time is usually the deciding constraint, and it is also the easiest one to evaluate honestly, because it can be measured on your own line with your own boards.
An inline inspection station must complete its inspection within the takt of the line, or it becomes the bottleneck. In practice this means the inspection time per board has to fit inside the time the slowest upstream machine allows. Where inspection is inherently slower than that, for example detailed X-ray measurement of dense area array packages, either the station is placed offline or the inspection is applied to a sample.
Conveyor buffers upstream of an inspection station smooth short bursts and keep the line moving during brief waits. They do not change average throughput. If inspection is slower than the line on average, buffers fill, upstream machines block, and output settles at the inspection rate. Buffers are useful for absorbing variation, not for covering a structural mismatch.
An offline station runs in parallel with the line rather than in series with it, so its cycle time does not subtract from output at all. For inspection methods that take longer than the line allows, or for products with low enough volume that the line is rarely the constraint, offline inspection is often the only configuration that makes economic sense.
Before choosing, measure the actual takt of the line with the product you care about, and measure the inspection time per board for that product on the candidate machine. Compare them. This single comparison settles most configuration questions quickly and prevents the common mistake of installing an inline station that quietly caps output.
Where inline inspection is marginally too slow, running two lanes, or splitting inspection across two machines, can restore balance. This adds capital and floor space but preserves both coverage and throughput. It is a legitimate answer for high-volume products where sampling is not acceptable.
The two filters sometimes point in opposite directions: the risk analysis says inspect every board, and the timing says the method is too slow to sit inline. That conflict is common with X-ray and with detailed three-dimensional measurement. The resolution is usually a split rather than a compromise. Run the fast checkpoints inline at full coverage, and place the slow, high value method offline with a defined sampling rate, tightening to full coverage only for the products or periods where risk justifies it. Recording that reasoning in the plan is what makes the decision defensible later.
The second filter is how much of the product you need to see. This is a risk question rather than a speed question, and it often overrides the throughput answer.
Inline inspection sees every board, so a process excursion is detected on the board where it starts rather than several hundred boards later. That immediacy is what makes inline inspection worth its cost on high-volume or high-consequence product: the feedback delay between cause and detection approaches zero, and the number of suspect boards is small.
Sampling accepts a detection delay. If a process drifts between samples, the boards produced in between carry the defect, and the size of that exposure depends on the sampling interval and the volume produced during it. For low-severity defects this exposure is an acceptable trade; for safety related, high warranty cost, or contractually reported defects it usually is not.
If you sample, write the rate into the plan along with the logic behind it: risk classification, historical stability, and the volume produced between samples. A sampling rate that nobody can justify is the one that gets argued about after an escape. Sampling should also tighten automatically after any process change, and widen again only once stability is demonstrated.
X-ray illustrates the tension sharply. Hidden joint defects are invisible to every other method, yet X-ray is often slower than the line and expensive to place inline. The common resolution is inline X-ray for high-reliability or high-density product, and offline X-ray for sampled audit, new product validation, and failure analysis. Configurations such as the X-7100 support the offline audit role, while the X-7900 and the X-9200 address higher resolution and automated requirements.
Solder paste inspection and pre-reflow AOI are the checkpoints where full coverage pays back most reliably, because they catch defects before value is added. Running these inline is often the first configuration upgrade a line makes, and it typically reduces downstream rework more than adding inspection further along the line.
Coverage is sometimes decided for you. Customers in automotive, medical, aerospace, and other high-reliability segments may specify inspection and record requirements contractually, and industry acceptance standards such as IPC-A-610 for assembled boards and J-STD-001 for soldering requirements define what an acceptable joint is, with published guidance available from IPC. Where a contract or a standard requires per-board evidence, sampled offline inspection cannot satisfy it regardless of how well it is run, and that requirement alone settles the configuration question for the affected checkpoints.
Configuration also determines how boards move, how much floor space the inspection function consumes, and how many mechanical interfaces have to work reliably every shift.
Every manual transfer is an opportunity for handling damage, bent boards, dropped assemblies, and electrostatic events that are hard to trace later. Inline inspection removes transfers entirely for the boards it covers. Offline inspection adds at least two, out of the line and back into it, and the more often that cycle repeats, the more handling risk accumulates.
Inline machines occupy line length, which is usually the scarcer resource in an existing building. Offline machines occupy floor area beside the line and need clearance for an operator, a magazine rack, and the flow of boards to and from the line. Neither is free, and the right choice depends on whether your constraint is line length or surrounding floor space.
An inline station has to agree with its neighbours on board width, transfer height, direction, and the handshake signals that govern when a board is released. These interfaces are mature and widely supported in SMT equipment, but they still have to be commissioned, and any mismatch in height or width handling becomes a daily irritation. Offline equipment has no such interface, which is a real advantage in mixed-vintage lines.
Lines that run a wide range of board sizes, thicknesses, and panel formats place more demands on an inline station's handling. Very large, very thin, or heavily populated boards may be easier and safer to inspect offline. Where mix is extreme, keeping one inspection method offline avoids constraining the whole line to the worst case handling scenario.
Adding inline inspection to a running line means physically inserting a station, which requires line length, conveyor adjustments, and a shutdown window. Offline equipment can be installed in parallel without touching the line at all, which is why many first inspections are bought offline and moved inline later, once volume justifies the disruption.
Double-sided assemblies complicate configuration, because the board passes through printing, placement, and reflow twice, and each pass has its own checkpoints. Inline inspection handles the repetition automatically on each pass, which is one of its quieter advantages. Offline inspection doubles the handling per board and requires someone to keep track of which side has been verified. Where a line also carries a through-hole or DIP stage, an inline position dedicated to that stage, such as a DIP online inspection system like the AI-4026, keeps the whole sequence under one conveyor.
What happens after a fail is where configuration differences become visible in daily work, and it is frequently overlooked during evaluation.
An inline reject can be routed automatically to a fail lane, an NG buffer, or a marker, and the line keeps running. An offline finding depends on the operator to route the board. Both work, but only the first is automatic, and automatic routing is what makes repair statistics trustworthy, because no human decision stands between detection and the record.
Every repaired board should be re-inspected before it rejoins the flow. Inline configurations make this natural, because the board simply passes the station again. Offline configurations require a deliberate step, and that step is the one most often skipped when the shift is busy. If you run offline inspection, make re-inspection a documented part of the repair procedure, not an informal habit.
A line with both inline and offline inspection produces results in two places. Those results have to land in one record with one defect taxonomy and one board identifier, or cross-stage analysis becomes manual. Confirm that the offline machine can export to the same system as the inline one before you buy, because reconciling two vocabularies later is far more expensive than checking first.
Customers in regulated or high-reliability segments increasingly ask for per-board inspection records. Inline inspection generates those records automatically as a byproduct of running. Offline inspection can too, but only if the workflow includes recording the board identifier at the station, which adds a step per board and therefore a cost per board.
When a defect escapes, the first question is which boards are affected. Complete inline records answer it immediately. Sampled offline records answer it only within the sampled population and leave an unsampled interval to reason about. The cost of that ambiguity should be weighed against the saving from not inspecting every board.
Whichever configuration is chosen, the pass or fail decision has to trace back to a defined acceptance standard rather than to an individual operator's judgement. Inline systems enforce this by construction, because the program applies the same limits to every board. Offline systems rely on the operator applying the same interpretation every time, which makes written criteria, reference images, and periodic calibration between operators more important. Where inspection records are used as evidence, that consistency is the part auditors look at first.
Configuration determines who does the work of inspection and how often they have to do it, which over a year is usually a larger cost than the capital difference.
Every job needs a program, and that effort is largely independent of whether the machine is inline or offline. What differs is how often the effort is repeated. High-mix lines running many jobs per week multiply program preparation time, and on those lines an offline station can be more forgiving, because program development and debugging do not hold up production.
Offline inspection consumes operator time in proportion to board count: carry, load, start, review, and return. On a high-volume product this cost is continuous and significant. On a low-volume product it is negligible. Calculate the labour per board honestly and compare it against the capital and integration cost of going inline.
Both configurations need someone who can distinguish a real defect from a false call, but inline stations need that person on call rather than present. An inline station with a poorly tuned program can stop a line or flood a fail lane while nobody is watching. Plan the response: who receives the alarm, what they do, and how quickly.
Inline inspection is part of the line, so any changeover that touches it affects line availability. Offline inspection can be set up while the line runs something else, which is a genuine advantage where changeovers are frequent and short runs are the norm.
If inspection depends on an operator, every shift needs that operator. Where labour is constrained, or where night shifts run unattended, inline inspection is the only way to keep coverage. This is often the practical argument that settles the decision in factories running three shifts.
Configuration changes what travels with a job. An inline checkpoint carries its program, its limits, and its results inside the line control, so the next run of that job inherits them automatically. An offline checkpoint needs someone to load the correct program revision and confirm it, which is a documented step rather than an automatic one. On high-mix lines with many jobs in circulation, the discipline of program revision control matters as much offline as online, and it is worth designing the job release process around it.
Capital cost is the least interesting part of the comparison, and it is the part most often used to decide. The useful comparison is total cost per inspected board over the life of the equipment.
Inline configurations carry the machine cost plus integration: conveyor interfaces, line modification, commissioning time, and the shutdown window. They also consume line length, which has an opportunity cost if the building is full. In exchange, marginal cost per board is close to zero once the program is stable.
Offline configurations generally avoid integration cost and installation disruption, and the same machine can serve several lines, which improves utilisation. In exchange, they carry recurring labour per board, more handling risk, and slower feedback. Whether that trade is favourable depends almost entirely on volume and on severity of the defects involved.
A single offline machine can support several lines, several products, and both audit and failure analysis work, so its capital is amortised broadly. An inline machine serves one line. In low-volume or multi-line environments, that utilisation difference can dominate the entire comparison.
The largest cost in the comparison is usually the one not written down: the cost of defects that reach the customer because inspection was sampled rather than complete. Warranty handling, returns logistics, root cause investigation, and relationship damage are all real costs of the offline choice, and they should be estimated on the same page as the capital figures.
A staged path is often the most defensible: start with offline inspection for validation and sampled audit, move the highest-leverage checkpoint inline first, usually solder paste inspection, and add further inline coverage as volume and process maturity justify it. This keeps early capital low while building toward complete coverage where it pays back.
Phasing also reduces technical risk. Starting offline lets a team learn a detection method, tune programs against real boards, and build the defect taxonomy before any line is modified. By the time the same method is moved inline, the programs are mature and the false call behaviour is understood, which is exactly when integration disruption is easiest to justify. The staged path costs a little more in total capital than buying inline immediately, and considerably less in disruption and rework.
Very few factories are purely inline or purely offline. The practical work is deciding, checkpoint by checkpoint, which configuration fits, and writing that decision down.
A common configuration runs SPI and post-reflow AOI inline because both are fast and both catch defects early, keeps X-ray offline for sampled audit and failure analysis unless the product demands full coverage, runs first article inspection at changeover where the setup risk lives, and places electrical test according to mix, with online ICT for stable product and offline ICT or flying probe for high-mix work.
Go through the checkpoint list and mark each one inline or offline using four questions, remembering that the method itself has already been settled by the defect it has to catch, as set out in our guide to matching SMT inspection methods to defect risks: is full coverage required by risk, does inspection time fit the takt, is there line length and integration budget available, and is there an operator available on every shift. The answers differ by checkpoint, and they should produce a mixed answer on most lines.
Write down, for each candidate checkpoint, the product volume, the defect severity, the required coverage, the measured inspection time, the available line length, the staffing plan per shift, and the data system the results must feed. With those seven items, a configuration decision is straightforward. Without them, it is a guess dressed up as a preference.
Whichever configuration you choose, validate it on your own product before committing: run your boards, at your takt, with your programs, and count real defects and real false calls. Supplier demonstrations are useful, but the measurement that matters is made with your boards in your factory. The evaluation criteria behind that trial are covered in our guide on how to choose SMT inspection equipment.
Configuration decisions expire. A product that justified offline sampled audit at introduction may justify inline coverage once it ramps, and a product that justified inline coverage may drop to sampled audit when it enters end-of-life decline. Revisit the assignment when volume or mix shifts materially rather than treating the installation decision as permanent.
As a summary: inline fits high volume, high severity, fast methods, unattended shifts, and tight feedback loops. Offline fits low volume, slow methods, high mix, multiple lines sharing one machine, new product introduction, and failure analysis. Most lines contain both, and the framework above is what tells you which checkpoint goes where.
Once the decision is made, the equipment follows: inline AOI platforms such as the AI-5146 and the dual side AI-5146W sit inside the flow, the inverted camera AI-4540 addresses through-hole stages inline, and offline AOI covers sampled audit, new product work and repair verification. The full range of vision inspection equipment and the AOI inspection solutions from I.C.T include both configurations so the checkpoints can be matched to the line rather than the other way around.
No. Inline inspection gives complete coverage and immediate feedback, but it costs more to integrate, consumes line length, and must keep pace with takt. Offline inspection suits low volume, slow methods, high mix, shared use across lines, and new product introduction.
Only in an offline configuration. An offline inspection machine can be shared by multiple lines and used for audit, validation, and failure analysis, which spreads its cost. An inline machine is dedicated to the line it is built into.
Not necessarily. It means sampled rather than complete coverage and slower feedback. With a justified sampling rate, disciplined re-inspection after repair, and good records, offline inspection can be appropriate for low-risk product. It is harder to defend for high-consequence defects.
Usually solder paste inspection after the printer, because paste volume predicts many downstream defects and catching print faults early avoids rework entirely. Pre-reflow AOI is the other strong candidate where placement errors are a dominant risk.
Compare inspection time with line takt and weigh the severity of hidden joint defects. If X-ray fits the takt and the product carries dense area array content or high reliability requirements, inline automated X-ray is justified. Otherwise offline X-ray for sampled audit and failure analysis is the common arrangement.
Labour per board in the offline case and integration plus line length in the inline case. Both are routinely underestimated, and both recur, while the capital difference is paid once.