中文

Where to Place AOI: Before or After Reflow, 2D or 3D, in 5 Stages from Listing Defects and Choosing a Location to Covering What It Misses

Which station AOI (automated optical inspection) belongs at, and whether to use 2D or 3D, starts with where the defects you want to catch are created. Public sources say a location after placement and before reflow catches missing parts, offset, wrong polarity and bent leads, and is an easy place to fix them; a location after reflow catches most soldering defects; and 2D images carry no height information, so defects such as lifted leads and lifted components are harder to judge, while 3D can measure height. There are also things AOI misses: solder spikes that do not connect to another lead, solder balls that are not between leads, cracked components, and locations hidden by tall parts. In the same study, more than half the defects caught by electrical testing were ones AOI could not catch.

This article covers which station AOI goes at on an SMT line, what 2D and 3D can each catch, and how to cover what AOI misses. AOI equipment selection and brand comparisons, the details of solder paste inspection (SPI), and X-ray inspection methods are outside its scope. It has five stages:

  • Stage 1 | List the defects to catch and where they are created: which defects printing, placement and reflow each create.
  • Stage 2 | Choose the location: what you catch after printing, after placement and before reflow, after reflow, before wave soldering and before electrical test.
  • Stage 3 | Choose 2D or 3D: whether you need height depends on whether your defects need height to be judged.
  • Stage 4 | Cover what AOI misses: solder spikes, solder balls, cracked components, hidden locations, joints that cannot be seen, electrical problems.
  • Stage 5 | Implementation and tuning: building programs, reducing false calls, re-tuning when the process changes, running in parallel with manual inspection for a while.

Each stage covers how to do it and what public sources say, plus how JIAHON can help where relevant. Sources, exceptions and this article’s synthesis are in the notes at the end. This describes the general case; individual situations may differ.

Figure 1. Five stages for deciding where to place AOI (illustration). (1) List defects — top defects, ranked by type; where they are created: printing, placement, reflow; whether height is needed: lifting, coplanarity. Goal: a map of defects to stations. (2) Choose location — before reflow: missing parts, offset, polarity; after reflow: most soldering defects; inline or offline: fast response or more flexibility. Goal: the inspection station location. (3) 2D or 3D — need height: choose 3D; present or not, right or wrong: 2D is sometimes enough; 3D must deal with shadowing and warpage, which the vendor says it has handled. Goal: the inspection method. (4) Cover what it misses — solder spikes and balls not connecting leads or not between leads; hidden and invisible: tall parts, joints under components; electrical problems: keep in-circuit test. Goal: a checklist to cover the gaps. (5) Implement and tune — build programs from a component library; reduce false calls by tuning per product; run alongside visual inspection to build confidence. Goal: a stable inspection program.

Stage 1: List the defects to catch and where they are created

How to do it:

  • Classify past defects: rank the defects and customer returns of recent months by type and take the top few (this article’s synthesis).
  • Map them to the station that creates them: too much, too little or no solder paste comes from the printing station; missing parts, offset, wrong polarity and bent leads mostly from the placement station; tombstoning, shorts, insufficient or excess solder, non-wetting and voids become visible after reflow (Source 1’s classification).
  • See where each defect is easy to fix: the Source 3 case describes the pre-reflow location as an easy place to fix (Source 3); at that point the parts are not yet soldered, while finding the defect after reflow means rework (this article’s synthesis).
  • Which defects need height to be judged: lifted leads, lifted components and lifted pads are hard to judge from a flat image (Source 2).

What public sources say: a paper presented at an IPC conference by an author from an electronics manufacturing services company (2002, reviewing two years of AOI use) divides the possible AOI locations into five and lists what each catches: after printing, excess, insufficient or missing paste; after placement and before reflow, missing parts, offset, wrong polarity and bent leads; after reflow, missing parts, skew, tombstoning, polarity, bent leads, shorts, insufficient or excess solder, non-wetting and voids (Notes, Table 1).

How JIAHON can help:

  • For defining the problem clearly and ranking priorities with risk priority numbers before bringing in equipment, JIAHON’s published article “Should You Bring In New Equipment? JIAHON Completes the Evaluation Report for You” walks through the method using a solder paste inspection machine as the example. It also notes that when you rely on sampling, by the time the post-reflow AOI or test station catches the problem, a batch of defects has already gone out.

Stage 2: Choose the location

How to do it:

  • After placement, before reflow: catches missing parts, offset, wrong polarity and bent leads (Source 1). The Source 3 case describes this as an easy place to fix, and uses the inspection measurement data to calibrate the placement machine (Source 3).
  • After reflow: catches most soldering defects as well as placement defects. In that study, total coverage at this location was 80% to 100% across products, with a lower bound of 74% for soldering (Source 1).
  • Before wave soldering, before electrical test: before wave soldering, you can check whether through-hole leads are present. After wave soldering and before electrical test, you can see through-hole leads plus the post-reflow items, but both sides must be inspected, it may become a line bottleneck, it cannot catch SMT defects at an earlier step, and it is not well suited to statistical process control (Source 1).
  • Inline or offline: inline means less handling and faster response; offline can inspect boards from different stations, the inspection flow can be adjusted at any time, and with several lines a product can run on any of them (Source 1).
  • Solder paste problems: stop them first with post-print solder paste inspection (SPI), dividing the work with AOI (this article’s synthesis; JIAHON has a published article on evaluating a solder paste inspection machine).

Figure 2. Possible AOI locations and what each catches (illustration, based on Sources 1 and 3). The line runs printing → placement → reflow → through-hole and wave soldering → electrical test. After printing: too much, too little or no paste. After placement, before reflow: missing parts, offset, polarity, bent leads — easy to fix at this point. After reflow: tombstoning, shorts, solder volume, non-wetting, voids. Before electrical test: through-hole leads; both sides must be inspected and it may become a bottleneck.

What public sources say: the IPC conference paper compared several locations and chose after reflow. It says this location catches SMT defects at an earlier step and was not a bottleneck on their line, with boards found defective sent to an offline repair station; the author concluded this strikes a balance between quality and cost. Placing it after wave soldering and before electrical test means inspecting both sides, may create a bottleneck, and does not allow effective statistical process control of common SMT defects. Inline inspection means less handling and faster response; offline inspection can serve several stations and gives flexibility across lines. A contract manufacturer implementation case in an electronics manufacturing magazine describes placing component measurement before reflow, giving four reasons: one is that this is an easy place to fix; the others are that placement is the main cause of random defects, that the pre-reflow system is easier to program with fewer false calls, and that the products contain many BGAs. This manufacturer used the inspection results to trace placement causes such as misplaced board supports and bent nozzles, used the measurement data to calibrate the placement machine, and stopped the line to deal with it when quality fell below 90% or the same type of error appeared on three boards in a row (Notes, Tables 1 and 3).

An example of inspection in the wrong place: on a module with very small passive components, printing was checked by measuring paste thickness at just a few points on the strip, with no real-time feedback per pad. If placement was offset, polarity reversed or a part missing, it could no longer be seen once the shield can went on, and could not be reworked; in continuous production, defects would pile up in large numbers. The report listed printing and placement problems station by station and recommended adding SPI between printing and placement, and AOI between placement machines. The lesson: put inspection where defects can still be seen and still be fixed — before the shield can goes on.

Stage 3: Choose 2D or 3D

How to do it:

  • Whether you need height: if your main defects are lifted leads, lifted components or coplanarity, which need height to be judged, consider 3D (Source 2).
  • Fine pitch, small parts: 2D judges lifted leads by the area of the joint profile, which becomes less reliable as parts get smaller and pitch finer (Source 2).
  • Present or not, right or wrong: if you just need to judge whether a part or material is there and in the right position, 2D is sometimes enough (this article’s synthesis).
  • What 3D has to deal with: shadowing, measurement range and false data caused by board warpage are problems that must be handled when developing 3D measurement; the equipment maker’s article says these have been handled or solved (Source 2).
  • False calls: with flat images, it is harder to judge whether a suspect point is a real defect, so false calls are more common and skilled people are needed to re-inspect. Source 2 says 3D can reduce false calls, which is the equipment maker’s own claim (Source 2).

What public sources say: an article in an electronics manufacturing magazine by an author from a 3D AOI equipment maker (2010) says grayscale 2D images have no depth information, so lifted leads, lifted components and lifted pads are hard to judge; 2D judges lifted leads by the area of the joint profile, which becomes less reliable as parts get smaller and pitch finer; 3D measures the height of components, leads and board features and can check coplanarity, lifted leads and lifted components; and developing 3D measurement requires dealing with shadowing, measurement range and false data from board warpage, which the article says have been handled or solved. The article promotes the company’s own 3D products; this article takes from it just the explanation of how 2D and 3D differ (Notes, Table 2).

Stage 4: Cover what AOI misses

How to do it:

  • Defects AOI misses: solder spikes that do not connect to another lead, solder balls that are not between leads, components standing on their side (which need very fine tuning and may still not be judged), and cracked or damaged components; some machines also cannot recognize the component part number (Source 1).
  • Hidden locations: depending on the component layout, some tall parts may hide small parts next to them (Source 1). When laying out the board, you can list these locations and arrange a separate check (this article’s synthesis).
  • Joints that cannot be seen: joints such as BGA and QFN joints sit under the component, where optical inspection cannot see them, so X-ray or other methods are needed. For how to inspect BGA joints, see JIAHON’s published article “How to Inspect BGA Solder Joints: X-ray, Dye-and-Pry or Cross-Section, in 5 Stages” (this article’s synthesis).
  • Do not drop electrical testing: in that study, 58% to 61% of the defects caught by in-circuit test were ones AOI could not catch. The author concluded that dropping in-circuit test would let these defects reach the customer (Source 1).

Figure 3. What AOI misses needs separate coverage (illustration). Solder spikes and balls — spikes not connecting leads, balls not between leads: visual inspection, microscope. Cracked components — cracked or damaged parts: visual inspection, microscope. Hidden — tall parts may hide small parts next to them: list them at layout and check separately. Invisible joints — joints under components: X-ray or other methods. Electrical problems — in-circuit test still catches many: do not drop electrical testing.

What public sources say: the IPC conference paper says AOI cannot catch solder spikes that do not connect to other leads, solder balls not between SMT leads, components on their side (which need very fine tuning and may not be judged as defects), or cracked or damaged components, and that some machines cannot identify components; depending on layout, some tall parts may hide smaller ones; and of the defects found by in-circuit test, 58% to 61% could not be detected by AOI (Notes, Table 1).

How JIAHON can help:

  • For choosing between X-ray, dye-and-pry and cross-section for BGA joints, see “How to Inspect BGA Solder Joints” above.
  • For solder spikes, solder balls and cracked components that AOI misses, a 7-inch digital microscope can be used to photograph and confirm them during visual inspection (the product page lists 1 to 1,200x continuous zoom, an adjustable lens angle to avoid glare, and uses including confirming joint defects such as cold joints, bridging and solder balls).

Stage 5: Implementation and tuning

How to do it:

  • Building programs takes time: a new board needs a component library; once the library is built, setting up new boards is much faster (Source 3).
  • Reducing false calls: in that case, false calls came from components with dark leads and bright pads with cutback paste (the source’s term). They were brought down by building a separate library for each product and adding a multivariate classifier to define component images (Source 3).
  • Re-tune when the process changes: process changes such as a new solder paste, stencil or reflow oven all require AOI to be fine-tuned again (Source 1).
  • Run alongside manual inspection for a while: to build confidence in AOI’s judgments, first run several batches of boards through AOI and cross-check them with manual visual inspection, confirming that it catches the defects within its coverage (Source 1).
  • Evaluate with numbers: whether the equipment keeps up with cycle time, whether repeat measurements are stable, and how long the investment takes to pay back — follow the steps for an equipment evaluation (the method in JIAHON’s published article).

What public sources say: the IPC conference paper says developing and fine-tuning programs takes engineering time; process changes such as a new solder paste, stencil or reflow oven all require re-tuning; and to build confidence in AOI’s judgments, several batches of boards should be run through AOI and cross-checked by visual inspection to confirm the system catches all defects within its coverage. The magazine case says building a component library for a new board took about four hours, or under thirty minutes with an existing library, and a product changeover took about three minutes; after several systems were installed and new products and components were added, false calls exceeded 1%, caused by components with dark leads and bright pads with cutback paste; after building a separate library for each product and adding a multivariate classifier to define component images, they fell below 0.5% (Notes, Table 3).

Related reading

To cover the related process steps, continue with these articles:

  • “Should You Bring In New Equipment? JIAHON Completes the Evaluation Report for You”: the four steps of an equipment evaluation report
  • “How to Inspect BGA Solder Joints: X-ray, Dye-and-Pry or Cross-Section, in 5 Stages”: choosing between X-ray, dye-and-pry and cross-section for BGA joints

Notes: sources and exceptions

These notes are for readers who want to check the sources; you can follow the steps above without them. Where your product specifications, your customer’s acceptance criteria or the relevant standards differ from this article, they take precedence.

Sources for each stage

SourceWhere used, and source
Source 1Stages 1, 2, 4 and 5: what each of the five locations catches, trade-offs between locations, inline and offline, what AOI misses, tall parts hiding others, the share still caught by in-circuit test, re-tuning and parallel running. A paper presented at an IPC conference by an author from an electronics manufacturing services company (2002, reviewing two years of experience)
Source 2Stages 1 and 3: no depth information in 2D, unreliability at fine pitch, 3D height measurement, problems 3D must handle, false calls. An article in an electronics manufacturing magazine by an author from a 3D AOI equipment maker (vendor material, 2010)
Source 3Stages 1, 2 and 5: pre-reflow is easy to fix, tracing placement causes, time to build libraries, reducing false calls. A contract manufacturer implementation case in an electronics manufacturing magazine (vendor material; both authors list a company as their contact, whose business the page does not describe; no reliable date on the page)
JIAHON material“How JIAHON can help” and related reading: JIAHON’s published articles and the digital microscope product page

Table 1 | What each location catches and misses

Location or itemWhat public sources say
After printingExcess, insufficient or missing paste (Source 1)
After placement, before reflowMissing parts, offset, wrong polarity, bent leads (Source 1); easy to fix (Source 3)
After reflowMissing parts, skew, tombstoning, polarity, bent leads, shorts, solder volume, non-wetting, voids; in the study, total coverage 80% to 100% across products, with a lower bound of 74% for soldering (Source 1)
Before wave solderingWhether through-hole leads are present (Source 1)
Before electrical testThrough-hole leads plus post-reflow items; both sides inspected, may become a bottleneck, not well suited to SMT statistical process control (Source 1)
MissesSpikes not connecting leads, balls not between leads, components on their side, cracked or damaged components, some machines cannot recognize parts (Source 1)
HiddenDepending on layout, some tall parts may hide small parts (Source 1)
In-circuit test58% to 61% of the defects it finds could not be detected by AOI (Source 1)

Sources: see each row.

Table 2 | How 2D and 3D differ

ItemWhat public sources say
2DGrayscale image, no depth information (Source 2)
2D weaknessesLifted leads, lifted components and lifted pads are hard to judge; relies on joint profile area, unreliable at fine pitch (Source 2)
3DMeasures the height of components, leads and board features; can check coplanarity and lifting (Source 2)
What 3D must handleShadowing, measurement range, false data from board warpage; the equipment maker says these are handled or solved (Source 2)
False calls2D makes suspect points harder to judge and needs skilled re-inspection; that 3D reduces false calls is the equipment maker’s claim (Source 2)

Sources: see each row; Source 2 reflects a 3D equipment maker’s position.

Table 3 | Implementation and tuning

ItemWhat public sources say
Inline and offlineInline: less handling, faster response; offline: usable at several stations, flexible across lines (Source 1)
Re-tuningProcess changes such as a new solder paste, stencil or reflow oven all require fine re-tuning (Source 1)
Parallel runningRun several batches through AOI with visual cross-checks to confirm it catches defects within its coverage and build confidence (Source 1)
Building librariesAbout four hours for a new board, under thirty minutes with an existing library; product changeover about three minutes (Source 3)
False callsOver 1% after new products and components were added; below 0.5% after per-product libraries and a multivariate classifier (Source 3)

Sources: see each row; the figures describe those cases, not general rules.

Exceptions and limitations

  • What was and was not read: this article draws on 3 public sources (one IPC conference paper and two electronics manufacturing magazine articles) and JIAHON’s published articles. It did not read the IPC-A-610 acceptance clauses for each defect, or any standard for comparative testing of AOI equipment.
  • Age and perspective of the sources: Source 1 is a 2002 paper, and equipment then differed from today’s, so this article takes from it just the classification of locations and defect types. Source 2 is a 3D equipment maker’s article promoting its own products. Both authors of Source 3 list a company as their contact, the page does not describe its business, and there is no reliable date. The figures belong to those cases alone.
  • 2D is sometimes enough: “if you just need to judge presence and position, 2D is sometimes enough” is this article’s synthesis and does not mean other defects can be handled the same way.
  • Invisible joints: “joints under BGAs and QFNs cannot be seen optically” is this article’s synthesis; for X-ray methods, see JIAHON’s published article “How to Inspect BGA Solder Joints.”
  • This article’s synthesis: the defect classification in Stage 1, “parts not yet soldered” and “rework after reflow,” the SPI division of work in Stage 2, “2D is sometimes enough” in Stage 3, and listing hidden locations at layout and invisible joints in Stage 4 are this article’s synthesis. Every other statement has a matching passage in the sources or in JIAHON’s published articles.
  • Scope: this article covers which station AOI goes at, the differences between 2D and 3D, and how to cover what AOI misses. AOI equipment selection and brand comparisons, SPI details and X-ray inspection methods are outside its scope.
  • JIAHON’s pages: “How JIAHON can help” quotes published pages on the site (including product pages); specifications are as stated on the product page.

This article summarizes the general case. It is not a restatement of any standard, not an endorsement of any vendor, equipment maker or service provider, and not a promise about your product or process. Follow your product specifications, your customer’s acceptance criteria and the relevant standards, and your own validation; individual situations may differ.

Request a quote

Request a quote →

Back to top