Fig 1. International Rectifier adds key steps to its testing process to ensure its AU parts are suitable for automotive applications.
Fig 2. The normal distribution of RDS(on) for many wafer lots is used to set the minimum and maximum values shown on the datasheet.
Fig 3. The RDS(on) distribution of a single wafer can show the main population and the outlier parts.
Fig 4. During the DAPT process, designers can set part average test limits to remove outlier parts.
Your car is probably one of the most reliable pieces of equipment you’ll ever own. But even if it doesn’t run very well, given the environment it operates in as well as its long lifetime, it’s exceptionally reliable. Now compare your car to a state-of-the-art flat-screen television.
Both include lots of electronics, though their system requirements are very different. If the TV doesn’t work, it’s a nuisance. But if your car doesn’t work, it’s a huge inconvenience. And if something like your car’s brakes or steering fails while you’re driving it, the consequences can be serious or even deadly.
That’s why car system and component manufacturers go to greater lengths than many other “consumer goods” manufacturers to guarantee quality and reliability. So what can be done to ensure that the car is safe and dependable?
It starts at the component level, where two sets of qualifications are used to govern semiconductors: AEC-Q100 testing, which applies to integrated circuit products, and AEC-Q101, which covers discrete devices.
In both cases, parts must pass these qualifications before they’re released as “automotive grade.” The testing involved stresses the components in many different operational modes and conditions to check that the semiconductor and its package can remain within the stated performance even after many years in an automotive environment.
Testing Your Products
A discrete component such as a MOSFET or insulated gate bipolar transistor (IGBT) must meet several criteria to pass AEC-Q101 testing. For example, the electrical parameters must not exceed the datasheet limits. Also, the parameters must remain within ±20% of the initial reading of each test—with the exception of leakage limits, which aren’t to exceed 10 times the initial value for moisture tests and five times the initial value for all other tests.
To meet these qualification standards, it’s frequently necessary to make improvements through bills of materials, design, or testing, leading to a dedicated product line. International Rectifier offers a dedicated line of more than 300 automotive-grade components designated by the letters “AU” at the start of the part number (Fig. 1).
One such refinement for AU-grade components is the selection of a robust bill of materials, particularly the die attach, which is the material used to affix the semiconductor to the package. Typically, it’s solder-based. The AEC-Q101 testing calls for the part to be exposed to 1000 temperature cycles from –55°C to 150°C. Among many things, this test stresses the die attach material and checks its integrity after 1000 cycles.
As the drain connection of a power MOSFET is on the backside of the device and is made to the outside world through the die attach, measuring the RDS(on) is a good way to check for any degradation of the die attach material. The table shows the results for the RDS(on) shift of an “AU” or automotive-grade D2Pak power MOSFET that was put through temperature cycling testing.
|RDS(ON) Shift Of D2Pak Power MOSFET
After AECQ101 Temperature Cycling
|Number of temperature cycles||Average||Maximum|
The maximum shift of the parameters must be less than 20% after 1000 cycles to meet AEC-Q101 requirements. The part in this example exhibited a maximum shift of only 4.3% after 1000 cycles. At 3000 cycles, three times the criteria of AEC-Q101, the maximum shift was only 16.4%! This essentially proves that the part remains intact even under the harsh automotive environmental extremes and through the thermal effects of extensive power cycling.
The AEC qualification tests the component’s ability to withstand the physical environmental and electrical conditions in which it needs to operate. Yet longevity and low defectiveness are equally important features for automotive components.
Automotive system manufacturers expect component suppliers to ship components that will have zero failures per million devices, also known as 0 ppm. This might sound like tall order, but many steps can be taken to achieve this practical and achievable goal, such as dynamic part average testing (DPAT).
DPAT is a statistical method of removing parts with abnormal characteristics from the products that are sold to customers. For instance, assume we want to test the RDS(on) performance of a new power MOSFET. During the design cycle, many thousands of MOSFETs will be built over several manufacturing lots, or “wafer lots” as they’re known. If the RDS(on) of all of these devices is measured and plotted, it will form a normal distribution (Fig. 2).
Using the normal distribution, the upper and lower specification limits used in testing the part can be set, and these limits form the RDS(on) min and max values on any datasheet for the part. On a consumer- or industrial-grade part, these lower and upper specification limits will be used to determine if the part should pass or fail.
Consider the RDS(on) of a single wafer (Fig. 3). The distribution is still normal, but narrower, and a few rogue parts known as outliers aren’t in the main distribution. On face value, the outliers appear to be good parts since they fall inside the lower and upper specification limits. But since they aren’t in the main distribution, they must have some sort of defect compared to other parts on the same wafer.
Experience has shown that statically, these parts are more likely to have issues later in their operational life. They’re also at risk of having their parameters shift and move outside of the specification limits over time. To get to a lower ppm, the outliers should be removed. This is where DPAT comes into play.
DAPT probes all of the die on a given wafer. It also measures parametric values and plots distributions for each wafer. The DPAT algorithm is then run, and a unique set of part average test limits allows the parts in the main distribution to pass the test while the outliers are removed (Fig. 4). This method effectively removes bad parts at the time of production as well as the parts that have the potential to fail later in life, making it a positive step toward 0 ppm.
The importance of aiming for 0 ppm can’t be emphasized enough. Indeed, International Rectifier applies many more steps than the ones explained here to reach 0 ppm. For example, the passivation later can be put on MOSFETs. Also, hot and cold testing along with room temperature testing are undertaken on IC products.
Even after the DPAT process explained here, secondary screening steps like maverick die exclusion (MDE) and guard banding are carried out. These extra steps remove die from a wafer not only based on their own DPAT test result but also on the test results of adjacent die.
For instance, a part that tested well but was surrounded by bad die would be rejected. It’s like potentially buying a good house in a bad neighborhood. It could be a wise investment if the neighborhood is improving, but it’s ultimately a financial gamble—and gambling isn’t an option for automotive-grade products!
Benjamin Jackson is the product and business development manager for automotive MOSFET and DirectFET products at International Rectifier. He has a master’s degree in electronics and communications engineering with honors from the University of Bristol in the U.K.