September 1998
eMail the Editor
Criteria for Reliable High-Speed CSP Mounting
With csp consumption growing, electronics manufacturers are demanding placement modules that provide great flexibility and high throughput.By Gunter Schiebel, Siemens AG, Munich, Germany
Future SMT board designs will incorporate chip-size/chip-scale packages to an increasing degree. In fact, the annual growth rate of CSPs until the turn of the century, according to most forecasters, is by far the highest growth rate expected for any IC package type.Currently, about 50 different companies worldwide are offering or working on various CSP formats. CSPs achieved their initial acceptance in Japan for consumer products such as Sony camcorders. The format found its first significant applications outside of Japan only last year.
Due to the growing consumption figures of CSPs, particularly in telecommunications and computer applications, throughput is becoming a major issue. This paper addresses the various criteria for achieving high, practical placement rates and adequate placement accuracies with CSPs, using dedicated chipshooter systems.
The key criteria to be considered are the placement principle, component feeding, vision technique and possibly high-speed fluxing.
To achieve homogeneous, uncomplicated and the most economic SMD placement lines, electronics manufacturers are looking for placement modules that provide extreme flexibility and throughput.
Flexibility translates into the ability to deal with the complete SMD package-form range, including 0201 outline packages,and also covers die and flip-chip bonding processes. Placement modules should also accept all component supply formats.
One of the most critical features, which must be considered in context with the placement rate, is the placement accuracy. Physical laws, however, do not allow the achievement of extreme values in terms of speed and accuracy for certain technical approaches. As a result, different placement principles,in practice, show significantly different capabilities in the combination of their speed and accuracy.
Placement Principles
Pick-and-place clearly provides the best placement accuracy as long as the machine is equipped with a minimum number of placement heads. The quality of the key positioning axes, x, y and theta, is very important to the overall placement accuracy. With pick-and-place machines, the placement head is typically carried by an x-y gantry system.Within the placement head, the most important axis is the rotational axis, but the precision of the z-axis movement should not be neglected. In high performance pick-and-place systems, the z-axis movement control is typically equipped with a microprocessor, utilizing sensors for determining both the length of the vertical stroke and the required placement force.
Figure 1. Collect-and-place
A key advantage of the pick-and-place principle is that the precision placement head can be freely positioned in x and y. This positioning allows all specific lateral movements, such as those needed for waffle pack pick-up or multiple measurements on a component, over a stationary, upward-looking camera.
The most advanced pick-and-place systems today are specified with an x, y accuracy of ±20 µm based on a 4 Sigma quality level. The fundamental disadvantage of high accuracy pick-and-place systems, which normally provide only one high precision placement head, is the very limited placement rate. The rate of such systems is usually below 2000 cph, excluding any additional process activities such as fluxing.
Pure pick-and-place systems, which employ only one placement head, are rapidly disappearing. They are being replaced by more flexible systems.
One system, for example, provides (on a single gantry) a high precision pick-and-place head in combination with a multi-nozzle revolver head. Here the pick-and-place head is responsible for the placement of large BGA or QFP odd-shaped components and very challenging fine-pitch flip-chips.
The high-speed jobs, with smaller and less demanding components (in terms of placement accuracy) are performed by the revolver (shooter) head. These "less demanding" jobs include CSPs down to a ball pitch of 0.5 mm (20 mil). The placement principle used is called "Collect and Pick and Place." (Figures 1 and 2).
The most advanced version of such a collect-and-place system (based on ball centering) can perform high-speed placement of CSPs with at least 5,500 cph with a placement accuracy of 60 µm based on 4 Sigma. This system also allows for high-speed flip-chip mounting. This is based on bump centering with bump diameters of 110 µm on bump pitches in the range of 200 µm.
Figure 2. Six-nozzle head used on the Siemens SIPLACE 80 F5 chop shooter
Traditional Chip Shooters
Traditional chip shooters (Figure 3) are normally defined by a horizontally rotating turret head, which simultaneously picks the components from a moving feeder bank and places them onto a moving PC board.Thus, very high theoretical placement rates in the range of 40,000 cph are enabled. For high-speed mounting of area array packages, classical chip shooters can only be used to a very limited degree.
Their limitations are:
- No component pick-up possible out of matrix trays (waffle packs).
- Placement accuracy is not sufficient for most CSP users. Typical values are clearly higher than 100 µm at 4 Sigma.
- Component fluxing is not realistically possible.
In the alternative collect-and-place shooter system (Figure 4), each of the two revolver heads is carried by an x, y gantry. Thus, the revolver heads, equipped with 12 nozzles, have random access to waffle packs or matrix trays.
In a dual-beam version, this system can achieve 90 µm/4 Sigma overall placement accuracy (including theta deviation) with 20,000 cph for the standard SMD package spectrum.
For area array packages, a dual beam shooter system, due to time-intensive ball/bump find algorithms, can operate with a placement rate of >11,000 cph in many cases.
If, instead of ball centering, outline centering is applied, the maximum placement rate will be 20,000 cph.
Placement Accuracy
Figure 3. The Principal of operation for traditional chop shooters.
Figure 4. Collect-and-place
Key determining factors for the required ball/bump placement accuracy for area array packages are ball count and package weight.
An interrelationship exists between these determining factors. For CSPs, the placement accuracy requirement is strongly relaxed, compared with leaded ICs (QFPs/SOs) with the same pitch. This is one of the CSP's key advantages.
The maximum acceptable placement error is equal to half of the PC board substrate pad diameter in the case of circular pads without a soldermask. Misplacement of the solder paste over half of the PC board pad diameter can occur, but a mechanical contact between the ball/bump and the pad will still take place. As a result, even though the solder paste is misregistered, perfect self-alignment is virtually guaranteed.
Placement Formula
Assume that the PC board pad diameter is normally slightly smaller than the ball diameter for CSPs with a ball diameter of 300 µm. Also assume a pad diameter of 250 µm. The placement accuracy requirement can then be calculated according to the following formula:The placement accuracy expectation in reality, however, is much higher. To achieve a good process capability index (cpk), users are demanding 4 Sigma placement accuracies-clearly better than 100 µm.
Vision Technique
The primary approach for many modern SMD placement systems employs machine vision techniques. These vision techniques generally consist of a combination of PC board and component vision systems.PC Board-Vision System
In meeting the sometimes extreme placement accuracy requirements (especially for and with flip-chips) of a modern assembly, the importance of PC board fiducial and inkspot recognition should not be underestimated. Global fiducial and inkspot determination can be very difficult, due, in part, to color and contrast conflicts.Fortunately, chip-scale formats like the µBGA® package and other area array CSPs, have lower placement accuracy demands, which enable the number of local fiducial readings to be reduced.
Memory Cards
Since a very typical CSP application will likely be small memory cards, a PWB panel technique is mandatory. To obtain maximum throughput within an SMD production line containing several placement modules, a feature like "whispering down the line" can be extremely beneficial. Whispering down the line means that only the first placement module has to spend time for the ink-spot recognition.The individual inkspot situation is "whispered" (transferred) to the subsequent modules. By using such methods within a line consisting of four modules, 21 seconds of precious machine time can be saved.
Inkspot Images
Typically within such a panel each defective substrate is marked with an inkspot. With a 15-up panel, the placement machine's downward-looking PC board camera must capture 15 inkspot images, which will take about 7 seconds.Component Vision Systems
To meet user demands based on widely varying materials and surface properties between flip-chips and other area array packages,extremely powerful component vision systems are required.Component vision system capabilities (and vision system capabilities in general) depend on both lighting (camera) technique and the algorithms applied by the evaluation unit.
Outline centering, using backlighting or laser-side illumination, should prove suitable for CSPs. However, there is a negative impact on placement quality because of CSP outline (substrate) tolerances.
To eliminate the need for extremely fine CSP substrate tolerances, ball centering is mandatory for most users. Vision ball centering is only possible with a front-lighted system.
Maximum recognition reliability and repeatability can only be achieved with the use of a sophisticated, flexible lighting technique employing various light sources. Each light source should have a specific light emission angle.
The near-perfect quality of the CSP image in Figure 5 was captured by the revolver-head camera of a Siemens SMD placement machine. Though sometimes routing structures are still slightly visible, they can be suppressed easily by specific ball location algorithms.
High-performance SMD placement systems, which have to deal with the complete package-form range, including fine-pitch flip-chips, must have at least two component cameras.
The flip-chip camera must have a different lighting approach and a much higher camera resolution (magnification) than a standard camera.
Orientation check
Another strong argument for ball centering is the orientation check (commonly called Pin 1 recognition) for area array packages).This is the only SMD placement machine vision feature that can reliably prevent placing these packages with the wrong orientation. The orientation check is automatically included in the placement process when ball centering is carried out with non-regular (non-symmetrical) ball arrays.
With flip-chip technology, the bump arrays are almost always non-symmetrical. This makes determining orientation easier and is an important pre-condition for high-yield, fully-automatic SMT production.
For most BGAs and fine-pitch BGAs and CSPs, orientation centering is not yet in place. However, Tessera's µBGA format is a nice exception to most surface-mounted packages in that respect.
Algorithms
Algorithms suitable for standard SMDs cannot readily be applied for ball centering on area array packages. Very complex, time consuming but interference-tolerant contour search methods are more advantageous.Despite the fact that ball robustness is one of the great advantages of the µBGA CSP, the capability for automated vision inspection of the solder balls is sometimes needed. With a powerful, flexible lighting approach and specific inspection algorithms, ball inspection with respect to deformation presence/absence is possible, but only to a limited degree.
One should be aware that the primary job of a component vision system is the precise and fast centering of the various packages. Fast optical centering is only possible with a single shot (no multiple measurements).
The large field-of-view results in a relatively coarse resolution which is contrary to the requirements of precise ball inspection.
This suggests that the results of a ball inspection of the SMD placement machine vision systems will continue to be relatively limited. Still, improvements in the placement machine's ball inspection capabilities can be achieved by taking several ball array images with different lighting. However, such an approach would slow down the placement process significantly and make programming much more complicated.
Furthermore, there is a strong conflict between the demands of full array ball inspection and a good placement rate. In most applications to achieve acceptable bump find calculation times, and thus high placement rates, only a few (five, for example) balls in each corner of the package should be programmed.
The only place to carry out a real ball inspection should be at the package manufacturer's site and the only equipment must be dedicated inspection systems, which typically cost several hundred thousand dollars.
Figure 6. Dip fluxing
Feeding
Feeding of BGAs and CSPs is not critical and is typically done with matrix trays or standard embossed tapes. It is worth remembering that only collect-and-place shooters work with matrix trays.When CSPs have to be used in applications with a wide mix of standard SMDs, reliability and yield can be a concern.
The critical process step is solder-paste printing. If the selected stencil thickness is too high, the solder paste destined for the pads at the CSP site might remain in the aperture. There are two ways to overcome this potential problem:
- Use special stencils that enable applying different solder paste thicknesses. The various thicknesses within a stencil can be accomplished either by step etching or by additive methods. Since special stencils are more costly and create some PWB layout limitations, their use is not very popular in the SMT industry.
- For reliable soldering and good positional stability during PC board transport-to and through the reflow oven-use CSP dip fluxing (Figure 6).
Additional times per placement cycle, depending on placement principle:
- Pure Pick and Place: Approx. 0.7 sec
- Collect and Place: Approx. 0.3 sec
Conclusion
CSPs face a very promising future for their performance and packaging density. However, to realize the full potential of CSPs and make them available for the entire scope of electronics production, further research and development is required to improve processes, materials, packaging and equipment.With respect to SMD placement equipment, development work will likely be concentrated on vision techniques, higher placement rates, improved accuracy and greater productivity.
Mr. Schiebel is a product manager for SMD placement machines in the Production and Logistics Systems Division at Siemens AG, Munich, Germany. He holds a Dipl-oma in power engineering. From 1971-1989, he was an electronics development engineer with the Siemens Automation Division. Readers may contact him at guenter.schiebel@mchrm.siemens.de or at +49.89.722.26581, fax +49.89.722.24260.