Graphene is a two-dimensional (2D) material, a single layer of carbon atoms tightly bound together in a honeycomb pattern (see Figure 1). It is one of the most conductive materials in the world, and any electronic device made from graphene will exhibit very little power loss.

Understandably, there is a considerable market for graphene-based components. One of Pickering’s customers has developed a technique whereby the material can be grown directly onto a semiconductor substrate, thus allowing traditional fabrication processes to be employed.

One of the early devices the company fabricated and continues to make is for magnetic field sensing applications, such as detecting position, speed, and direction of travel/rotation. The use of graphene means the sensors are highly responsive to magnetic fields, and they exhibit excellent linearity with respect to temperature.

Figure 1 – The Graphene Council describes graphene as the thinnest material known to man (so thin that it is considered to be 2D), the strongest material ever measured, the best conductor of electricity, an ideal material for sensors and electronic devices, and the best conductor of heat (even better than diamond which, like graphene, is also a carbon material).

Wafer Testing

The sensors are fabricated on 2-inch diameter wafers. Initially, the sensor die was relatively large, and the company was getting only 60 dies per wafer. Following a geometry shrink, the same wafer can now hold 1,200 dies. This is good news not only from an economy-of-scale perspective but also in terms of end applications, as the magnetic field sensors can be smaller. Specifically, the die can fit in a 3x3mm QFN package, whereas before, 10×10 QFN packages had to be used.

However, with smaller die (and 20 times more per wafer), testing became a challenge, and the company had to turn to automation. The first decision made was whether to buy an off-the-shelf semiconductor wafer test system or build one in-house. Off-the-shelf was quickly ruled out as early indicators pointed to an 18-month delivery date, far too long for the company’s ambitious goals, plus tests were still to be developed. Or, to put it another way, how could the company specify the requirements of an off-the-shelf automated test equipment (ATE) solution when they were still exploring how to test graphene structures?

It was decided to build the ATE in-house, using off-the-shelf subassemblies such as jigs, probes, switchgear, and test equipment. This would be a quicker, more adaptable solution. Also, as a growing company serving a market hungry for graphene-based devices, scalability would be key. The ATE would need to support expansion, which meant engaging with equipment suppliers with extensive portfolios and a commitment to open standards.

The Tests

A prototype ATE was built and working by the end of October 2022 and released into production in May 2023. The ATE has a test bed that is moved by stepper motors, allowing the company to navigate around a wafer. For a die test, four probes are brought into contact with the die, and the four signals feed into a high-density Reed relay matrix card (see Figure 2) supplied by Pickering Interfaces.

The company performs several tests (currently up to about 30) per die. For each, a stimulus (such as a current or voltage of a specific magnitude and polarity) is switched via the matrix card through to the probes at set times. The ATE can also apply a magnetic stimulus near the die. The same matrix card is used to divert signals from the probes to the appropriate measurement equipment.
The tests are controlled by a PC running in-house software. Many tests are for quality control purposes to confirm that the die’s characteristics are within the limits presented on the datasheets of the sensors shipped to customers. Other tests are for process improvements – i.e., evaluating the effectiveness of tweaks to the fabrication process – and real-time measurements/readings can be shared within the company.

The company’s process improvement tests are gated in that any given die needs to pass some basic electrical tests (which can take between 10 and 15 seconds) before being subjected to further tests (including applying a magnetic stimulus). This adds about another five seconds to the test time. However, a full set of tests with repeat readings made can take up to 50 seconds. Also, tests can be performed at pseudo-random locations on the wafer and depending on the results can be halted, allowed to continue, or reconfigured on the fly.

Understandably, if a full set of tests is to be applied to all 1,200 dies, the matrix card needs to be driven at speed if the overall test time for a wafer is to be reasonable. As mentioned, the reed relays have an operating time of just 0.5ms, although the company is not currently operating them at this speed. Indeed, the company’s software errs on the side of caution, and plenty of settling time is allowed to mitigate against the effects of switch bounce. But there is headroom if the company needs to use it.

In addition to the fast-operating time of the relays, another major benefit is that the reed switches used within the devices are instrumentation grade and have a very low contact resistance, which is the total resistance between the relay’s switch pins when the switch is closed. This is important because, to take accurate readings, the total in-line resistance between the probes and the measurement instrumentation must be as low as possible.

Summary

As is common in the semiconductor industry, repeatability and consistency are essential. In this respect, Pickering’s reed relays, present on the matrix card, can perform millions and even billions of operations with little if any degradation in performance.

This high life expectancy gave the customer immense peace of mind when it embarked on its automation journey, safe in the knowledge that speed would not be delivered at the expense of foregoing high accuracy when potentially performing tens of thousands of tests per wafer.

Also, the matrix card’s PCI-based architecture made integration into the ATE easy and enabled the sharing of real-time test results within the company, which is essential for monitoring process improvements.

 

Reproduced here with the kind permission of Pickering Interfaces