RFID Improves Throughput in Probe

A significant driver of wafer final test throughput is the availability of the right probe card for a particular lot of wafers. In facilities with hundreds or thousands of probe cards, keeping track of where they are and ensuring they are available and ready for use when needed is a formidable challenge. When a card is misidentified and used on the wrong lot/product, throughput drops because of the need to retest the wafers. This article describes a methodology of tracking, validating and managing probe cards using the same radio frequency identification (RFID) technology that has become the de facto standard for wafer-lot validation. Estimated improvements in throughput, rework and data accuracy are also provided as a comparison to manual or semi-manual (barcode) methods.

RFID, wafer-lot validation, throughput

The single most common use of RFID in semiconductor manufacturing is to automate the step of confirming that a wafer lot is at the correct tool for its intended next process step to ensure the wafers are not misprocessed. This also enables automatic recipe loading to match the specific step for that wafer lot. In some fabs, even the tool operator's certification or training level is confirmed prior to allowing them to operate the tool.

The goal of these validation steps is to minimize or eliminate rework, scrap and downtime, all of which have a direct impact on throughput. The value has clearly been recognized in the industry, as nearly all 300 mm load ports ship with RFID systems installed, and many 200 mm fabs are opting to retrofit their equipment with these same RFID capabilities. The added benefit to the 200 mm fab without automated stocker systems is the ability to rapidly find work in progress — another boost to throughput as well as equipment utilization.

Photo used with permission of FormFactor Inc., Copyright 2007.

Gaps in common probe card tracking systems

While most probe/test facilities employ some methodology of tracking and managing its pool of probe cards, the vast majority of these tracking systems leave much room for error. Typical systems include manual (hand-keyed or written) and semi-manual (barcode) probe card ID entry and tool-based automated systems, in addition to the RFID-based fully automated systems.

Manual systems require an operator to type the probe card's ID number and location into a terminal (or write it in a log) as it moves from place to place. Semi-manual systems also require the human action of picking up the barcode reader and pointing it at the probe card's barcode in the correct orientation. In addition, the technician must also read the ID of the location where the card is being placed, making this a two-step task. This methodology leaves open the possibility of error when the probe card is placed into the wrong carrier, with the wrong ID label on it, causing the probe card database to be updated with faulty data and, in turn, allowing wafers to be tested with the wrong probe card. This type of error precipitates a number of otherwise unnecessary activities, which of course equate to added costs and reduced throughput, such as wafer retest, probe card test and repair if it sustains damage when used on the wrong wafer product, prober/tester downtime, and the labor associated with manually searching for the correct probe card.

Tool-based automated systems, such as those that involve the installation of an EEPROM onto a probe card, solve many of these issues, but do not address the tracking of probe cards in the storage or card test areas. They also tend to be specific to the prober manufacturer, and are not readily standardized across the test floor.

Facilities relying on probe card management (PCM) systems based on these technologies continue to suffer suboptimal throughput and increased tester wait time, which equate to downtime for a multi-million-dollar piece of equipment.

The primary goal of this program at Test36 was to mitigate or remove these problems by addressing the human error that tended to be the common cause when the wrong probe card was used. An additional goal was to reduce the time and effort required to log the probe card movements from place to place.

Using the wrong probe card for a wafer lot is only one of several common misuses. Ideally, a probe card will have been cleaned, maintained and serviced immediately prior to use. Each fab has a different threshold for how many touchdowns are acceptable between servicings. Automatic counting of touchdowns, both total lifetime and since last service, enable the PCM to flag a probe card and pull it out of service at a preset touchdown limit. This ensures only those probe cards that meet the fab's "ready to use" requirements are allowed into production.

RFID solution

The PCM system used at Test36 consists of passive RFID tags for each of the probe cards and FOUPs; Asyst's AdvanTag RFID readers and antennas installed onto the prober/testers (for both probe card ID and FOUP ID); and AMD's own software application that sits on top of existing prober/tester communications software and interfaces with the RFID readers and the host (manufacturing execution systems [MESs]) system (Fig. 1). It identifies and validates each asset used in the test/probe step (tester, prober, probe card and FOUP/wafer).

1. The PCM system identifies and validates each asset used in the test/probe step (tester, prober, probe card and FOUP/wafer).

The tester workstation runs the PCM software and communicates to the prober via a general purpose interface bus to the RFID reader via RS-232 and the MES. The reader system is connected to two antennae. One of these is mounted in the test head and communicates with the RFID tag on the probe card (Fig. 2). The other antenna is mounted to the prober's load port and communicates to the RFID tag on the FOUP (Fig. 3).

2. One of two antennae is mounted in the test head and communicates with the RFID tag on the probe card.

The RFID tag on each probe card (Fig. 4) is programmed with the card's ID number. The tags on the FOUPs are similarly programmed with the FOUP ID number.

3. The FOUP’s RFID tag is mounted behind the white bracket.

The RFID reader has its own device IDs (one for each antenna) so that its location is known to the PCM system. All of these IDs are used by the MES as follows:

• Probe card ID is referenced in the MES database to the card's history, test results, touchdown counts, and the lots (specific IC products) it is designed for
• FOUP ID is referenced to the ID of the lot currently in the FOUP
• Device ID is the specific antenna location that corresponds to a tester (for probe card ID) or prober (for FOUP ID)

4. The RFID tag on each probe card is programmed with the card’s ID number.

With these data, the PCM and MESs know which probe card is at which tester and which FOUP is at which prober. That information is looked up in the database to determine which lot is in the FOUP and whether or not that probe card is the right one to run on that lot. The correct test program is then loaded automatically.

Reading the probe card and FOUP IDs is done automatically when each asset is put into its place on the tool, removing the need for an operator to enter any data manually. Human involvement is not needed, hence the activities associated with logging probe card movements also are not needed.

The functional steps are shown in Figure 5.

Step 1: The PCM system reads the FOUP ID, probe card ID, test program version and tester ID.

Step 2: The FOUP ID is used to get the right lot ID, device ID and wafer sort (WS) operation from the Inventory Table, which is automatically generated based on the fab production control system.

Step 3: The appropriate (valid) tester ID, probe card ID and prober set-up will be determined from the Golden Table (maintained by the test engineering staff) through the given device ID and WS operation from Step 2. Probe card validity is additionally checked based on touchdown counts, which are updated in the database by PCM.

Step 4: The data on the prober/tester will be checked against the valid combination from the Golden Table — if this is correct, the operator is allowed to continue.

5. Data from each asset is compared with the “Golden Table” prior to processing.

Conclusions

The RFID-based PCM system successfully removed the human error inherent in the daily operation of the prober/tester. Probe card and lot IDs are entered via RFID without a human step — no typing or barcode reading. The test program, probe card and set-ups are always validated prior to testing. Rework caused by test operation errors has been eliminated. Throughput has been improved by an estimated 2-3% because of the automated operation flow, in addition to unmeasured throughput improvements caused by the eliminated need for rework. And test floor control has become nearly "bulletproof" because of the accuracy of the lot ID and probe card ID data approaching 100%. Prober/tester uptime improvements were not measured because of the particular ways uptime and downtime are defined and captured at Test36. But uptime improvements were estimated to be in the 3% range. This improvement can be linked to improvements in other metrics, such as equipment availability and utilization.

*Based on "Improving Final Test Throughput Via RFID Tracking Of Probe Cards," by Mike Ingamells and Jens Kober, 2005 IEEE International Symposium on Semiconductor Manufacturing.