Basics of parametric measurement - What is parametric testing? Learn more about the principles of parametric testing (Chapter 1 of Measuring Parameters)

What is parametric testing? We will introduce all the various parametric tests in this article and the next few parametric measurement articles.

Principle of parameter testing

"As a branch of science, the central activity of engineering is the design of new devices, processes, and systems"
— Myron Tribus

Reference: * In 1999, HP was split and Agilent Technologies was established; in 2014, Agilent Technologies was split again and the independent department was renamed to its current name - Keysight Technologies. The original Agilent products mentioned in this article are also some of Keysight's products.

What is parametric testing?

What is parametric testing? This is an interesting, but perhaps controversial question. However, typical parametric testing involves electrical parameter testing and characterization of four major semiconductor devices, including resistors, diodes, transistors, and capacitors. This is not to say that parametric testing does not involve other types of devices, but the vast majority of parametric test structures can be classified into these categories or a combination of these categories.

Figure 1.1 — Parametric test design for the four basic device types

Most parametric tests involve either current-voltage (IV) measurements or capacitance-voltage (CV) measurements.

For many people, parametric testing means “DC” testing, but this is not accurate. Sure, it is possible to make measurements anywhere from milliseconds to seconds with source/measure units (SMUs), which is indeed “slow” by the standards of functional testers (which typically make measurements in the nanosecond or picosecond range). In recent years, the need for very fast parametric measurements has grown dramatically (single-point measurements of 1 µs at nanosecond data sampling rates). This has necessitated the development of new types of measurement modules (such as waveform generator/fast measurement units - WGFMUs) that meet this need. In the future, very fast IV measurements and pulsed IV measurements will continue to grow in importance as more sophisticated lithography techniques and new materials are adopted in device manufacturing.

A major subcategory of parametric testing is reliability testing. Reliability testing relies heavily on the well-known Arrhenius equation that describes the rate constant (K) of a chemical reaction:

Here: A is the prefactor

• Ea is the activation energy

• R is the ideal gas constant

• T is the temperature in Kelvin

In device reliability testing, high current and/or high voltage stress (greater than the normal operating conditions of the device) is usually applied to the device to reduce the value of the activation energy and increase the probability of the failure mechanism. Increasing the temperature can usually achieve the same purpose. Once the failure mechanism is generated, various mathematical and statistical techniques can be used to deduce the expected failure rate under normal operating conditions.

Why perform parametric testing?

The purpose of parametric testing is to determine the characteristics of a semiconductor process. In general, parametric testing includes the following three main areas: process development, device modeling, and process control.

Figure 1.2 — Parametric testing focuses on three main areas: process development, device modeling, and process control.

The first two are performed in the laboratory, R&D or pilot production environment, and the last one is performed in the manufacturing environment. The parameter measurement equipment used in different environments will obviously have different requirements. Process development and device modeling can use benchtop instruments, while production processes should use testers with high throughput.

Figure 1.3 - Parametric tester designed for production line use to optimize throughput

It is important to understand that parametric testing is almost never performed on the final product. Instead, it is performed on specific test structures that provide sufficient information about the process itself. Parametric testing is always performed directly on the semiconductor wafer. In production test, parametric test structures are sometimes located in the wafer's reticle rows or "streets" to minimize the wafer area occupied by the devices. Process development and reliability testing is usually performed on the parametric test structures rather than the entire wafer.

Figure 1.4 – To save valuable wafer area, parametric test structures are sometimes placed within the reticle rows (or “streets”) of the wafer.

Where to perform parameter testing

In production, parameter testing is usually performed after the wafer manufacturing process is completed (i.e., passivation has been performed) and before the electrical performance of the product die is sorted (electrical sorting).

Figure 1.5 — Parametric testing is performed after wafer fabrication and before product functional verification.

Each wafer in each batch is tested and the data is saved in a database. Obviously this is a huge amount of data, which can be processed into a variety of different formats using various software tools. One popular format is the wafer map, which uses different colors to represent different ranges of data values ​​on the wafer plotted with scalars.

Figure 1.6 — Example of a wafer map.

Conventional test structures placed in the reticle rows, or even insert test dies placed around the wafer, may not adequately characterize an advanced process. Due to their inherent complexity, advanced processes often require many more tests, and sometimes it is difficult to fit all the required test structures into the available test area. To maintain a reasonable lifespan for the probes on the probe card, the probes on the probe card are limited in physical size, which in turn limits the minimum size of the probe pad. This means that with each new process generation, the probe pad cannot shrink as the device feature size decreases.

Figure 1.7 – The probe pad size cannot be changed as design rules change, limiting the number of conventional test structures that can be placed on the wafer.

One solution is to use an array, as arrays allow the devices under test to share the probe pad, thereby increasing the device-to-probe ratio. Below is an example of this approach.

Figure 1.8 — An example of an addressable array scheme for parametric testing in production.

Compared with conventional device testing solutions, the throughput of addressable arrays is significantly improved. Although it can greatly shorten the test time, it requires the engineering effort of re-implementing all parameter test processes.

History of Parametric Instruments

The first instruments that could perform these parameter measurements were analog graphers. These instruments had limitations, however, notably that the "data" was displayed on a CRT screen. The only way to save the data was to take a picture of the screen display with a Polaroid camera, but this still did not give you actual numerical information.

1980s

In the early 1980s, Agilent Technologies* (then part of Hewlett-Packard) introduced the world's first digital parameter analyzer, the 4145A. The 4145A was the first instrument to combine four source/measure units in one instrument, complete with the software needed to integrate all the resources. The 4146A produced a plotter-like graph, but the 4145's graph included discrete points of digital data that could be transferred to other software for storage and analysis. This was a revolutionary product for the semiconductor industry, and the 4145A quickly replaced the plotter except for a few special applications. It was followed by an improved version, the 4145B, later in the 1980s.

*In 1999, Hewlett-Packard was split to form Agilent Technologies; in 2014, Agilent Technologies was split again and the independent department was renamed to its current name - Keysight Technologies.

Figure 1.9 — 4145B Semiconductor Parameter Analyzer

1990s

In the early 1990s, Agilent Technologies (then part of Hewlett-Packard) introduced the 4155 and 4156 semiconductor parameter analyzers. These were the 4145A/B instruments with a variety of new features added, including a color display and keyboard, automatic data analysis capability, pulse sweep capability, thinned sampling mode, DC and AC stress, and standby mode (to name just a few of the new features). There were multiple versions of this product ("A", "B", and "C"), with each successive version adding new features and measurement capabilities. In particular, the "C" version added quasi-static capacitance-voltage (QSCV) measurement capability and greatly improved the measurement capabilities of the voltage measurement unit (VMU).

Figure 1.10 — 4156C Precision Semiconductor Parameter Analyzer

2000s

As the 21st century dawned, it became clear that the requirements for parametric testing were becoming increasingly complex, requiring a complete solution that went beyond IV measurements. The solution had to be able to perform both CV and IV measurements, and also have the flexibility to add other measurement resources as future test requirements evolved. To meet these challenges, Agilent introduced the Keysight/Agilent B1500A semiconductor device analyzer and Agilent EasyEXPERT software in 2005.

The B1500A supports all aspects of parameter testing, from basic manual measurements to wafer test automation linked to a semi-automatic wafer prober. Because the B1500A uses the Microsoft® Windows® XP Professional operating system, it can be easily integrated into a variety of PC-based work environments. In addition, the familiar Windows graphical user interface (GUI) and convenient online help menu also minimize the need for instrument training. As the tool of choice for contemporary parameter measurements, the B1500A has replaced the 4155C and 4156C.

Figure 1.11 — B1500A Semiconductor Device Analyzer

The B1500A is a modular instrument that supports a variety of different module types. Modularity is a very important performance feature because parametric testing is still a very dynamic field, and new types of test requirements are constantly emerging. Parametric testing requires more than simple source/measurement unit resources, and complex tests including capacitance testing, high-speed testing, and fast pulse measurements are becoming more and more common. Modularity ensures that the measurement equipment will not become obsolete, and you have the ability to add new test capabilities to your parametric measurement equipment as future test needs evolve.