中文English

How to Measure Dispensed Adhesive Height? An Inline Inspection Case with Structured-Light 3D and 2D AOI – Jiahon Can Help

Process automation|Vision inspection case

To measure adhesive height accurately, first decide the measurement method: for height use 3D (this case uses DLP structured-light projection plus an industrial camera, accuracy ±0.1 mm), and to check only presence and area use 2D AOI; before the measurement system goes live, use GR&R to confirm repeatability and reproducibility.

In manufacturing quality control, one class of problems is especially tricky: you can see it but cannot measure it accurately; you can measure it but cannot put it on the production line. Inspecting the coating height of thermal adhesive is exactly such a problem.

Where is the problem?

After thermal adhesive (blue adhesive) is applied to a circuit board, its height must be within the specification range — too little adhesive and the cooling effect is insufficient; too much and it overflows during assembly, causing contamination. It sounds straightforward to judge, but automatically judging on the production line whether the adhesive height of every board is acceptable is harder than you might think.

1. The inherent error of measuring on the move

The traditional method is to let the product move along the conveyor and scan images as it travels. The problem is that conveyor transport can never be perfectly smooth, and slight up-and-down bouncing or shaking distorts the measurement data — however precise the lens, it cannot make up for this inherent error.

2. The trade-off between FOV and accuracy

The larger the measurement range, the lower the accuracy tends to be. The customer needed to cover several coated areas on a single product at the same time without sacrificing accuracy; this requires choosing the right system with the right settings, and cannot be solved by buying a standard unit and installing it as is.

Jiahon’s approach

After receiving the request, Jiahon did not recommend an off-the-shelf solution right away; it first evaluated the customer’s line on site to confirm several key conditions: the line’s dwell time, the PLC model (Mitsubishi), the actual product size, and the distribution of the adhesive.

After the evaluation, a combination of a DLP structured-light projection module and an industrial camera was selected, with these core specifications:

Measurement itemSpecification
FOV (scan range)220 × 140 mm
Measurement accuracy±0.1 mm
Repeatability (measured)Within 0.09 mm (90 μm)
Single image capture timeAbout 500 ms
Time to return resultsWithin 2 seconds

The key decision: static capture rather than scanning on the move. Images are captured during the product’s natural dwell time at the shooting position, removing at the root the measurement error caused by conveyor vibration, so that the accuracy figure truly reflects actual capability instead of being diluted by transport shake.

Delivered in two phases, not all at once

Jiahon split the whole project into two phases, rather than asking the customer to commit the full budget at once:

Phase 1: Capability verification (about 4 weeks)Phase 2: Inline automation (8–12 weeks)
Install the equipment on the customer’s line; complete standard setup and light shielding; run standard inspection on the blue adhesive; deliver a data interpretation and repeatability verification report; collect follow-up issues and fine-tuneConnect to the Mitsubishi PLC to complete I/O signal integration; trigger inspection automatically when the product arrives; display Pass/NG results in real time; extend coverage to areas not completed in Phase 1; complete overall automation verification

Why phases? They let the customer first verify feasibility with a smaller investment, and only after confirming it commit to the cost of automation — rather than betting on a solution not yet verified. This logic is more practical for both sides.

The same thinking, extended to white adhesive inspection

While the blue adhesive project was under way, the customer raised another need: there was also white adhesive on the same line — could it be inspected too?

The white adhesive challenge was completely different. The blue adhesive work was height measurement (3D), while the white adhesive needed presence judgment (2D AOI) — confirming whether there is adhesive at the specified positions and whether the area meets the standard. In addition, the white adhesive was spread over a wider area, right next to the board frame, with many interfering objects around it (empty pads, the housing, and so on), and the shooting angle was limited, so the blue adhesive solution could not be applied directly.

Jiahon redesigned the solution for the white adhesive, with this AOI workflow:

Step 1Step 2Step 3Step 4Step 5
Teach the inspection areas offlineAutomatically extract the red imageAnalyze the white adhesive areaJudge Pass / NGReturn the result to the line

A high-resolution camera (1.4 megapixels, able to distinguish 1 mm × 1 mm of white adhesive) was used, with a dual-camera module capturing the upper and lower halves, using the color contrast with the darker green PCB base to identify the white adhesive areas automatically. The whole process completes capture and judgment within 1.5 seconds, and supports dynamic capture (no need to stop the line).

This extended application was at first outside the scope of the two parties’ cooperation, but precisely because the blue adhesive project had built enough trust and technical foundation, there was room for the later cooperation.

On SMT / EMS and packaging lines, how adhesive volume is controlled

Measuring adhesive height inline is the final confirmation; a common practice on the line is to move adhesive volume control forward, to dispensing parameters and the first article. How several dispensing-related specifications in our engineering files put it:

  • Dispensed weight has a tolerance. A dispensing monitoring specification for flip-chip underfill requires dispensed weight to be within ±10% for targets of 30 mg and above, and within ±15% for targets below 30 mg; the dispensing temperature of the hot plate must be within ±5 °C of target.

  • Measure the drop volume before production. A die coating specification requires measuring “the total drop volume of 10 units” once before starting the line, then repeating it 2 more times; production starts only when the three agree.

  • Track adhesive thickness with control charts. A die attach specification lists uncured adhesive thickness (the average height at the four corners of the die, minus the die thickness) as an X-bar R control chart item, sampling 2 dies each time.

  • Look at fillet height for forming. The acceptance criterion in the same underfill specification is: the adhesive fillet height at the center of each of the four sides of the die must be greater than “half the die height + the bump gap,” and a CSP with no fillet is rejected.

These items control “the amount dispensed,” while inline 3D measurement controls “the final shape”; with both in place, when adhesive height is abnormal you can tell whether the dispenser’s output has changed or the height of the substrate or components has changed.

The measurement system itself must also be verified: the NIST/SEMATECH Engineering Statistics Handbook lists the gauge R&R study as a method for assessing the error of gauges used in production (Section 2.4), while the automotive supply chain commonly uses AIAG’s Measurement Systems Analysis (MSA) manual (MSA-4); for the method and interpretation, see “What is GRR? Gauge repeatability and reproducibility“.

What this case shows

Jiahon’s role in this project was not that of an equipment dealer, but closer to a partner in solving a process problem: taking part throughout, from defining the problem, selection and evaluation, and custom software development, to phased delivery and verification.

The technical routes of the two cases (3D measurement of blue adhesive, AOI presence judgment of white adhesive) are completely different, but the problem-solving logic is the same: think the problem through first, then choose the right tool, not the other way around.

FAQ

Q1: Should adhesive height be measured with a laser or structured light?

Both are optical 3D measurement. Structured light captures the whole field of view at once (in this case FOV 220×140 mm, single capture about 500 ms) and suits measuring many points at the same time; laser triangulation suits single points or line scans. Which to choose depends on the number of points to measure, the cycle time, and the accuracy.

Q2: How much adhesive counts as acceptable?

It depends on the adhesive type and the product specification. Specifications commonly state a dispensed weight tolerance (for example ±10% / ±15%) plus forming acceptance criteria (for example fillet height, extent of overflow); in this case the height specification of the blue adhesive is defined by the customer’s product.

Q3: What verification is needed once inline measurement is installed?

First do GR&R to confirm that the proportion of measurement variation to the tolerance is within an acceptable range; then check regularly with standard parts of known height to confirm the measured values have not drifted.

Similar vision inspection needs on your line?

Measurement, judgment, or automation integration — you are welcome to start from a description of the problem, and Jiahon will look at how it can be solved with you.

Contact Jiahon

Products mentioned in this article|specifications and quotation

© Jiahon Group|Jiahon Enterprise Group|jiahon.com.tw

Back to top