Wafer-Level Test and Burn-in (WLTBI)

Wafer-level Test and Burn-in (WLB)

Wafer-level Test and Burn-in (WLTBI) refers to the process of subjecting semiconductor devices to electrical testing and burn-in while they are still in wafer form. Burn-in is a temperature/bias reliability stress test used in detecting and screening out potential early life failures.

WLTBI usually employs a wafer prober to supply the necessary electrical excitation to all the die on the wafer through hundred or thousands of ultrathin probing needles that land on the bond pads, balls, or bumps on the die. The required die temperature elevation, on the other hand, is achieved by the wafer prober through a built-in hot plate that heats up the wafer to the correct junction temperature.

Wafer-level testing and burn-in is applicable not only to: 1) devices sold as bare die, which are also referred to as 'known good die' or 'KGD'; and 2) wafer-level packaged devices; but to 3) devices intended for conventional packaging as well. In the third case, WLTBI is performed as a prescreen, so that only the parts that passed WLTBI will undergo back-end processing, i.e., assembly and final test.

The ideal semiconductor manufacturing scenario is to come up with a process that does everything at wafer level, but its prohibitive costs for now does not make it viable for all applications just yet. Once perfected, however, an integrated wafer-level packaging, wafer-level electrical testing, and wafer-level burn-in will streamline the over-all semiconductor manufacturing process to a large degree, resulting in great cost savings and much shorter cycle times.

After all, wafer-level packaging, which is basically just an extension of the traditional wafer fabrication process to provide each die on the wafer with a means of interconnecting to the outside world, would eliminate the need for a separate IC packaging/assembly line. In a similar fashion, wafer-level testing and wafer-level burn-in would eliminate the need for separate equipment for testing and burn-in, since both of them may be performed on a wafer using the same basic methodology and set-up.

This is not so in the case of individually packaged IC's, whose electrical testing and burn-in require different equipment in different areas on the production floor. Conventional electrical testing uses expensive automated test equipment (ATE) on the test floor while conventional burn-in requires burn-in ovens that are kept in their own burn-in areas due to the large amounts of heat that they radiate.

Still, the basic philosophies used in testing and burning in individual IC's are the same as those used for wafer-level test and burn-in. Both electrical testing and burn-in need a means of supplying the devices under test (DUT) with electrical bias and excitation, whether it's done at wafer level or at package level. The difference lies in the method of delivering the required electrical bias and excitation to the devices.

During electrical testing of individual IC's, electrical bias and excitation are provided by the ATE to the DUT by mechanically contacting its leads. In conventional burn-in of individual IC's, the units are placed on burn-in boards which in turn are inserted inside burn-in ovens. The burn-in ovens provide the electrical bias and excitation needed by the devices during burn-in through these burn-in boards.

During wafer-level testing and burn-in, however, the electrical bias and excitation required by the devices are delivered directly to the interconnection points (the bond pads or the solder balls/bumps over the bond pads) of each die on the wafer. This can be achieved in a variety of ways, some of which are discussed in the next page.

Achieving Full Wafer Electrical Contact

The challenge in any wafer-level testing and burn-in process is being able to use existing wafer probing technology to contact all the operation-essential pads of all the die on the wafer at the same time. This is referred to as full-wafer or whole-wafer contact technology. The capability to do so will allow the burn-in process to be conducted to the entire wafer in one operation.

Once electrical contacts have been made, wafer-level devices may already be subjected to the same testing methodology as what their individually packaged counterparts normally receive. In electrical testing, this may mean subjecting the DUT to a sequence of test blocks, each of which forces a certain set of voltage and/or current conditions to the DUT and measures the corresponding current/voltage/timing response of the DUT against specifications.

Burn-in, on the other hand, places the DUT in an electrically stressful condition over a specified amount of time. Stressful electrical conditions include operating the device at maximum power dissipation, continuous dynamic switching of the inputs, application of high reverse bias voltages, and the like.

In an article by Dan Inbar and Mark Murin of M-Systems (source: Semiconductor International, 8/1/2004), the formidability of achieving whole-wafer contact with today's wafers was explained using a simple example: if a typical wafer has 500 die, with each die containing 40 functional pads, then 20,000 probing points are needed to properly activate all of these die on the wafer during burn-in. Cramming all of these probe needles onto a single 6" wafer at the same time without allowing any of them to come into contact is indeed challenging.

Full-wafer contact systems currently employ three different methods or technologies: 1) the probe-per-pad method; 2) the sacrificial metal method; and 3) the built-in test/burn-in method.

The example above wherein each pad of each die on the wafer is directly contacted by an ultra-thin contact pin or needle of a wafer probing system so that electrical testing may be performed by the test equipment pertains to the probe-per-pad method. Needless to say, the challenge presented by this method is coming up with a proper design for an extremely dense array of probes.

In the sacrificial metal method, a thin layer of metal is deposited over the entire wafer in patterns that connect together the equivalent bond pads of groups of die on the wafer, so that a reduced number of probe needles may be used to excite all the die on the wafer. After the WLTBI process is completed, this sacrificial layer is etched away from the wafer. The main drawback of this method is the need for extra wafer fab steps to deposit and remove the sacrificial metal layer.

The built-in test/burn-in method involves the application of Design-for-Test (DFT) philosophy in the development of new products. Here, a new device would incorporate an additional special circuit on the die that would facilitate self-testing and/or self-burn-in using a relatively smaller number of probes. Such a circuit might employ serial I/O (to reduce the number of I/O probes needed) and a built-in test/burn-in subsystem. Wafers of this new product may then undergo full-wafer contact probing using a much smaller number of probes.

Challenges in Wafer Level Test and Burn-in

Aside from the high up-front costs of developing and setting up the equipment, especially if full-wafer contact technology is involved, another serious challenge posed by WLTBI is the achievement of highly reliable and excellent electrical contact between each of the probes and its corresponding bond or test pad (or bump) on the die.

Poor contact or loss of contact in the middle of test or burn-in may result in a multitude of problems: over-rejection, insufficient burn-in, and even electrical overstress (EOS). Contact failure involving even just a single pad or bump of the device will cause the test or burn-in to fail. A wafer probing system with excellent and reliable contact capability will eliminate yield losses due to contact failures - a necessity in the ever-competitive semiconductor industry.

Ensuring high contact integrity and reliability for a large number of probe tips is not easy though. At the very least, it entails a sound maintenance routine that consists of monitoring probe tip life, replacing worn-out tips, and continuous tip-to-pad alignment checks and realignment.

Achieving the bandwidth required by electrical testing and burn-in of high-speed devices is also another consideration that needs to be addressed by an engineer setting up WLTBI capability. Excellent engineering design and material selection for the probe needles to be used are a 'must' if high bandwidth test and burn-in capability is desired.

The initial steps toward viable wafer-level test and burn-in systems have already been taken. Still, the journey toward industry-standard WLTBI methods and equipment will be long and arduous. Obstacles in the way of standardized WLTBI processes include the large diversity of wafer-level packaging solutions; the continuous reduction in wafer size and die interconnection pitches; and the wide availability of conventional test and burn-in solutions for applications that are not yet ready for WLTBI.

Just the same, many companies will try to be at the forefront of WLTBI technology because, to most of them, an integrated wafer-level manufacturing approach would make the best sense for the future of semiconductor manufacturing.

Figure 1. Example of a Handler for

Wafer-Level Burn-in

See Also: Electrical Testing; Burn-in; Probe/Trim;

Wafer-Level Packaging; IC Manufacturing

HOME