Becoming a semiconductor producer is a tricky process these days. In formerly low-cost industrial areas, wages and energy prices rose, while capital spending climbed. At the same time, competition is heating up, with a plethora of new businesses entering the market in recent years. Industry players are understandably anxious about these changes, and they have been pursuing a record amount of M&A activity in an attempt to capitalize on the next wave of productivity growth.

Semiconductor manufacturing is divided into two phases: "front end" and "back end". Back-end semiconductor fabrication refers to the fabrication operations after all the circuits are formed on the wafer. Revolutionary technology has been created by combining extraordinary accuracy and precision with enormous throughput.

Many operations in back-end semiconductor production employ servo drives because of their excellent performance and repeatability, which is required for high-end semiconductor processing.

Most back-end factories located in emerging countries are yet to use Industry 4.0 technologies in their key operations, including wafer dicing, assembly, testing and packaging of individual semiconductors. Many of these factories are still struggling to implement the lean methods common in front-end factories. Even when back-end manufacturers reap some benefits from lean initiatives, they often struggle to maintain progress.

The relevance of back-end activities in semiconductor production continues to increase in the face of growing consumer demand and increasing industry competitiveness. More efficient tools are needed to assist machine setup and batch scheduling decisions to achieve short cycle times, high throughput and high utilization while improving due date performance.

List backend tool process

Wafer inspection

Optical wafer inspection looks for defects that could cause problems in the final product. Defects and annoyances as small as 30nm can be detected and effectively as small as 10nm. Electron beam detection overcomes the limitations of optical detection and is accurate to sub-3 nm resolution. E-beam inspection can identify the tiniest faults compared to optical inspection, but has lower throughput. As flaws and annoyances are identified, they are mapped and corrected or avoided.

Wafer Test/Wafer Probe

These chips are tested the first time throughout semiconductor manufacturing to ensure they function as intended. Functional checks are performed while the chip is still on the wafer, using a test fixture with pins that make contact with the circuitry on the chip's surface. The chip's signal response is sent and measured by the probe. If possible, faulty chips are repaired; otherwise, they are destroyed after the dicing process.

Dicing wafer

In this back-end semiconductor manufacturing process, the finished wafer is diced into individual chips. Mechanical sawing and laser cutting are two ways of automation. Cutting saws use circular cutting blades to cut molds to sizes from 35mm to 0.1mm for mechanical sawing. The die is then transferred to the die bonding process using die handling equipment.

Servo motion is suitable for aligning the dicing saw and wafer as well as adjusting the dicing blade.

chip bonding

Individual chips are too small and fragile to handle individually. They must be protected and there must be an easy way to electrically connect to the chip. The process of bonding the bare die to the substrate is called die bonding or die attach.

In the next process, the substrate will serve as the interface between the tiny size of the chip and the large-scale electronic processing. It will also serve as the basis for a PC board protection chip package.

wire connector

Wire bonding uses thin gold wires to connect each pad on the die to a corresponding pad on the substrate after the die is bonded. This connects the silicon chip inside the chip container to the pins on the outside through electrical connections. Wire bonding is used in traditional chip packages such as the dual in-line package (DIP), which has a characteristic black oblong rectangle with silver pins protruding like bug legs, and the PLCC package, which has conductors on all four sides.

Wire bonding machines operate at breakneck speeds to maintain the high number of connections required for each chip. In fact, this is one of our most bandwidth-intensive applications.

Flip Chip/Solder Ball

Flip chips are mounted "backwards" as a modern alternative to wire bonding. As a result, the term "flip chip" was coined. Unlike wire bonding, where the wires are attached around the edge of the chip, an array of "bumps" is created on the chip's surface. These bumps serve as connectors between the chip and the surrounding container. Here are some benefits of flip chip technology:

    A better connection to the chip, rather than wire bonding, adds extra length, capacitance, and inductance, all of which slow down the signal speed.
    Since the entire chip is exposed, not just the border, more connection sites are accessible.
    increase production speed
    Overall package size is small.

encapsulation

At the completion of the back-end semiconductor manufacturing process, the bonded chip and frame are sealed using a molding plastic compound or by attaching a sealing cap. Silicon chips are now ready for use in the electronics industry.

How to optimize backend tools?

Unleash the full potential of your workforce

Operator Contact Time Employees spend 30% to 50% of all back-end plant work time touching materials or running machinery. Workers often sit idle for the remainder of the workday while they wait for machines to complete their manufacturing cycles. Even when the production line is not running at full capacity, the ratio of workers to machines is consistent, which increases the duration of inactive workers.

Standard lean practices, such as varying worker-to-machine ratios based on operator contact time or employing flexible staffing to ensure shop floor personnel numbers are sufficient for a plant's current capacity, have helped some back-end manufacturers increase labor productivity. These initiatives have yielded some benefits, but they have been difficult to maintain, meaning back-end production remains labor-intensive.

Improve quality without delaying the production line

Engineering teams must study machine data and communicate with colleagues on the production line to identify specific production steps that are lost in the event of peaks or losses in production or unexpected quality issues in back-end facilities. However, engineers may only collect data once a week, long after the problem has arisen, making it more difficult to pinpoint the root cause.

Engineers may need to interview line employees for information, and workers may recall some basic data about tool settings or other operating environments, which can cause delays.

Establishing dedicated yield and quality improvement teams and daily Lean "meetings" may be a preferable approach. These structured discussions can help engineers grasp issues such as output stability and unpredictability so they can make improvements.

Consider throughput in a more proficient way

Most backend plants rely on absolute metrics of uptime equipment being available or unavailable, ignoring subtle results such as small outages that don’t result in a complete shutdown when evaluating OEE. Additionally, backend manufacturers use manual procedures to track production losses, which only reveal broad patterns over time. These high-level conclusions do not give engineers a comprehensive understanding of what is causing production problems, making it difficult to develop improvement strategies.

In order to solve these problems, some return to the fundamentals of Lean is required. For example, a manufacturer might form a continuous improvement team to prioritize and pinpoint the source of throughput bottlenecks. Many organizations have these teams, although they don't always appear in backend factories.

Consider a simple idea: machines could be equipped with sensors to track significant events affecting OEE, such as production failures or equipment failures. Operators will then enter contextual data through a touchscreen interface, saving time on manual data entry and providing engineers with a higher level of detail.

All in all, the semiconductor business is a leader in data collection; the problem is that companies only use some of the data they get. For the first time, advanced technology can help manufacturers tap into their vast knowledge bases to provide the specific, actionable insights needed to develop solutions.

Additionally, Industrial Revolution 4.0 tools automate many time-consuming processes that are now done manually in back-end factories. Together, these enhancements help managers execute lean initiatives faster and more efficiently, with some organizations seeing meaningful cost, throughput, and quality benefits within months.

Back-end factories that integrate smart manufacturing technologies could stand out in the highly competitive semiconductor industry, outperforming those employing more traditional lean methods.