Using AOI to verify IPC compliance
Dr. Malachy Rice, Agilent Technologies, Dublin, Ireland -- Test & Measurement World, 11/1/2004
Automated optical inspection (AOI) platforms generally employ one of two design approaches: multiple low-resolution cameras mounted at various angles around a component, or a single high-resolution camera mounted above. The use of an AOI system on a production line can help a manufacturer ensure its electronic devices comply with the IPC's A-610 guidelines for printed-circuit assembly.
The IPC (www.ipc.org) established A-610 to help manufacturers achieve the highest possible SMT production quality. The standard specifies three classes of electronic devices, depending on how mission critical the final application is.
![]() |
| Fig. 1 The dimensions labeled here are defined as critical parameters in IPC A-610. |
Tight component positional tolerances minimize the risk of an electrical short to an adjacent component. For a QFP, the risk is greatest in the left-right direction (defined as A in Figure 1), while for a chip, the risk is greatest in the axis parallel to the chip body (defined as B).
From the example in Table 1, consider an 0201 chip and a 256-pin QFP with a 32-mm body size. If the 0201 pads are 400 µm wide (defined as P) and 740 µm long, then for Class 3, the 0201 chip device has a tolerance of 125 µm in the A direction:
(component width – pad width)/2 +0.25 * component width
In the B direction, it has a tolerance of 70 µm:
(component length – pad length)/2
For the QFP, the component lead width will be 200 µm. Assume the pad width is 300 µm. There is no definition of positional tolerances in the B direction. In the A direction, the positional tolerance for Class 3 is 100 µm.
To measure offsets accurately and repeatably without false failures, an AOI system must feature even tighter tolerances. Generally accepted guidelines dictate that the tolerance of a measurement-system be less than 10% of the tolerance of the unit the system is trying to measure. Therefore, reliably determining Class 3 positional compliance requires an AOI system measurement tolerance of no more than 7 µm for the example 0201 chips and 10 µm for fine-pitch QFPs.
To meet these specifications, AOI system designers must pay attention to vision-algorithm repeatability, x-y table repeatability, mechanical-frame rigidity, and even factory-floor stability. Single-camera, high-resolution AOI systems designed for pre-reflow component measurement can usually cope with these specifications better than multiple low-resolution camera AOI systems designed primarily for post-reflow joint-inspection measurements.
Post-reflow AOI systems assume a joint position identical to the position of the corresponding pads. Because the system gets this information from CAD data, it need not be calculated, permitting looser positional-tolerance design constraints.
Angled camera AOI systems experience even greater limitations when inspecting PCBs that are slightly warped. When looking at an object from the edge, these systems will interpret any vertical deformation as a horizontal offset. Although software algorithms can compensate, the accuracy and repeatability of multiple camera systems depend on the sophistication of this compensation, and generally they cannot match the performance of orthogonal camera solutions.
One final factor that affects positional tolerance repeatability is the camera's pixel size. When measuring the
|
|
| Fig. 2 Accurately detecting small devices requires a small pixel size. The top image--taken with a camera with 15-µm pixels--is clearer than the bottom image, which was taken with a camera with 50-µm pixels. |
As you can see, the 15-µm pixels—which are available in higher-resolution cameras—create a clearer image. Single-camera AOI systems can afford to include the best high-resolution cameras. Cost considerations generally do not allow multiple-camera systems that luxury.
Checking jointsIPC guidelines include solder-joint specifications for a range of devices from chips to fine-pitch QFPs. Assuming that the AOI system knows the component's width and thickness from the parts library and knows its pad width from the CAD software, it can determine whether the joint width, joint length, and joint height, defined in Table 1 as C, D, and F, respectively, meet the standard.
For chips, either angled or orthogonal AOI systems can measure 2-D joint width C and length D. Both types of AOI systems have full visual access around the joint.
But if the joint is extremely small, inexpensive AOI systems can have trouble determining whether an 0201 device even has a joint—and would find it impossible to handle the more difficult challenge of measuring it. So, as with positional tolerances, an AOI system's ability to measure critical joint dimensions depends heavily on its measurement repeatability, which again favors the use of a single, high-resolution camera system.
Any AOI system would have difficulty measuringjoint height F. The IPC guidelines specify that joints should reach at least one quarter of the way up the chip body. Measurement of such a joint would require a 3-D laser triangulation system like those used for solder-paste inspection. Neither AOI approach could perform this task.
IPC joint specifications for QFPs require the measurement of the same three critical parameters as with chips—length, width, and height. Either angled or orthogonal camera AOI systems can see toe fillet width C, but measuring this dimension again favors high-resolution single cameras. Neither type of system can see IC heel fillets, however, so manufacturers will need x-ray inspection to measure the width of these joints.
Side fillets on IC joints present a different problem. The IPC defines the side-fillet length as a critical measure of a good quality joint. Because QFP pins reside close together, positioning the camera and adequately illuminating the side fillet are major considerations in AOI system design.
The optimum solution is to position the camera directly above the joints and thereby eliminate occlusion when device pitch becomes very small. Near-vertical light sources also minimize the effect of shadows from adjacent pins. Therefore, orthogonal camera systems could measure side fillet length D for even the finest-pitch devices. Angled camera systems could see only the side fillet nearest the camera and only if it is not in shadow.
The IPC also describes good joints with visual models meant as guides to help repair operators understand a good solder fillet's shape characteristics. A few of today's advanced AOI systems can show a repair operator an exact 3-D model of a joint, allowing the operator to compare it with a stored picture of a known-good joint.
Numerous techniques allow recreating a 3-D view of a scene from a series of 2-D images. These techniques rely on the fact that any 3-D object has different surface reflections and different shadows when viewed from the same spot and illuminated from different directions.
This technology is still in its infancy, but its capabilities are already proving useful to operators. These advanced 3-D techniques rely on high-resolution imaging—a single camera mounted a fixed distance from the PCB with the device lit from all sides. Low-resolution angled cameras cannot perform the same function.
| Feature | Label | Chip | QFP |
| Side overhang max. | A | Classes 1,2: A < 0.5 W or 0.5 P | Classes 1,2: A < 0.5 W (or 0.5 mm) |
| Class 3: A < 0.25 W or 0.25 P | Class 3: A < 0.25 W (or 0.5 mm) | ||
| End overhang max. | B | Classes 1,2,3: B not permitted | Class 1,2,3: B < electrical clearance |
| Joint width min. | C | Classes 1,2: C > 0.5 W or 0.5 P | Class 1,2: C > 0.5 W |
| Class 3: C > 0.75 W or 0.75 P | Class 3: C > 0.75 W | ||
| Joint length min. | D | Classes 1,2,3: D > 0.5 W or 0.5 P | Class 1: D > W (or 0.5 mm) |
| Class 2,3: D > 0.75 L or W | |||
| Fillet height min. | F | Class 3: F > G + 0.25 H | Class 2: F = G + 0.5 T |
| Class 3: F = G +T |
| Author Information |
| Malachy Rice holds a primary degree in electrical and electronic engineering from the University of Edinburgh, as well as a PhD in signal processing for telecommunications from the Queen's University of Belfast. He is currently a product marketing engineer for Agilent Technologies' optical inspection systems. |



















