Current trends in the design of semiconductor devices are forcing design and test engineers to reevaluate the status quo and to realize that it is no longer possible to use traditional design rules and test methods with new semiconductor systems. More and more frequently it is becoming necessary to implement methods involving frequencies in the gigahertz range, which can be daunting for engineers who up until now have had nothing to do with RF or microwaves.


That is why AccuraCV,™ now available in the SussCal® Professional calibration and measurement software suite, has been introduced. It addresses the challenges resulting from shrinking device sizes by making accurate impedance measurements of physically small elements. Having the capability to make such measurements is becoming increasingly important—for instance, the International Technology Roadmap for Semiconductors (ITRS) reports that in 2012 the thickness of gate oxides, characterized using impedance measurements, is predicted to be about half as thick as it is today.

Impedance measurements are critical in process control, but traditional DC methods for making impedance measurements suffer inaccuracies when used on small components. As a solution, impedance characterization can be done using scattering (S)-parameter measurements at microwave frequencies. Thus, AccuraCV is an intuitive tool for optimizing the frequency of impedance measurements that will provide the most accurate results. In addition, it can be used to optimize device design by reducing the costs incurred during process control.

Current Status

There are several different conventional methods for measuring device under test (DUT) impedance. The most commonly used are those available from commercial impedance analyzers (based on the measurement of the DUT current over the applied voltage). These are widely used in the semiconductor industry and provide an accurate impedance measurement at frequencies up to 110 MHz. However, for very small devices, extracting the modeling parameters of a DUT is becoming difficult in this frequency range, and unwanted effects such as current leakage significantly reduce measurement accuracy.

To overcome this problem, additional effort must be put into the measurement procedure. This includes using more complicated three-, four- or five-element DUT equivalent circuits, in which the parasitic elements—those elements used to describe effects such as current leakage—must be characterized accurately. This reduces the practical application range of conventional impedance measurement methods for characterization of advanced semiconductor components. These limitations can be overcome by selecting S-parameter-based impedance measurement techniques and increasing the measurement frequency to the microwave range.

Microwave Measurement Methods

The S-parameters (reflection and transmission coefficients) of the DUT can be obtained with the help of a vector network analyzer (VNA). The main advantage of a VNA-based measurement system is that it is relatively simple to measure S-parameters in a very wide frequency range in one sweep, from some hundreds of megahertz to beyond 110 GHz. A very simple relationship between the reflection coefficient ΓDUT of the DUT and its impedance ZDUT allows the measurement software to extract the desired model parameters easily and at any frequency:

where

Z0 = measurement system reference impedance

However, the reflection coefficient is a relative parameter. It is derived in a measurement system with fixed reference impedance Z0, which is set at 50 Ω for most applications. It is simple to show that the constant error ΔΓ introduced in the reflection coefficient measurements due to the VNA measurement uncertainty leads to a nonlinear error function for measured DUT impedance. This fact limits the accuracy of the impedance characterization method based on the S-parameter measurements.

Accurate Modeling

AccuraCV was developed to meet the challenge of making accurate impedance measurements using S-parameters. It provides accurate modeling of the expected measurement error from the VNA-based measurement system. The mathematical model takes into account the estimated value of the measured impedance of the DUT as well as the measurement uncertainty of the system. As a result, the patent-pending AccuraCV algorithm calculates a value of the impedance measurement error over the specified frequency range.

Typically, most test elements can be measured at different frequencies. This requires the test engineer to find the optimum frequency for accurate impedance measurements either by trial and error or daunting mathematical calculations. With the help of this new tool, the optimal frequency can be found and the measurement error can be reduced to less than five percent, depending on the measurement setup and the DUT.

Figure 1 demonstrates the measurement results of microwave capacitance/voltage (C/V) characterization of the gate oxide capacitance of a next-generation semiconductor component optimized and analyzed with AccuraCV. To achieve the highest measurement accuracy for the region of smallest capacitance, the algorithm calculated the optimal measurement frequency as 1.3 GHz. At this frequency, the expected measurement error is reduced to two percent. Additionally, the capacitance extraction error is evaluated over the whole bias range and displayed as error bars.

Testing impedance of bias-dependent semiconductor components, such as transistors and varactors, the optimum frequency can be calculated for each bias point, increasing the measurement accuracy over the whole range of the impedance variation of the DUT.

Additionally, the AccuraCV algorithm optimizes test elements used in process monitoring during fabrication. The design of the verification element and its estimated electrical characteristics can be easily adjusted to the measurement equipment type and the uncertainty of the measurement method. As a result, the quality of process monitoring during production is increased.

It is recognized that the cost of testing increases proportionally with the test frequency. Therefore, from a cost-effectiveness perspective, the test should be kept at the lowest feasible frequency. By doing this, the new tool increases the effectiveness in device design, making it possible for test elements to be designed to optimize the required measurement frequency range.

For example, limiting the test frequencies at which process monitoring elements are tested to 6 GHz will reduce the cost of test because a less expensive measurement setup is required. AccuraCV easily calculates the required impedance of the test element at the specified frequency

(< 6 GHz) that will result in the least amount of measurement error. The test element can then be designed according to these specifications.

Conclusion

To decrease the manufacturing costs of semiconductor devices, more and more elements are being put on the wafer, decreasing the size of the components and increasing chip density. Due to the miniaturization of devices, S-parameter-based impedance measurement methods are becoming more commonplace at both the design and production phase. AccuraCV optimizes S-parameter-based methods by determining the test frequency that will significantly improve measurement accuracy.

The tool also finds the balance between the often competing requirements of decreasing the cost of test and increasing the test accuracy. From specifications calculated using the software, the test structures can be designed for accurate characterization at lower frequencies.

Last but not least, AccuraCV is integrated into the SussCal Professional calibration and measurement software suite, providing a full set of tools required for accurate S-parameter test at the engineer’s fingertips—from automated system calibration to measurement optimization. As such, the wizard-driven design of the software guarantees that a user with little or no RF or microwave measurement experience can make accurate measurements in a matter of minutes.

SUSS MicroTec Test Systems, Dresden, Germany
+49 35240 730,
www.suss.com.
RS No. 302