The semiconductor industry is always looking for new specialty materials, dielectric solutions and new device shapes to further reduce device size, and then further. For example, lateral and longitudinal heterostructures of 2D materials have led to new disruptive small-scale low-power electronic devices.
When it comes to producing accurate reports on the electrical characteristics of semiconductor devices, such as specialized NANO-FETs, researchers, scientists and engineers in the industry are faced with a common problem. This problem becomes even worse when it comes to proving that these parameters can actually be controlled in a simple, repeatable way.
4200A Electrical Parameter Characterization System Controlled from Touch Screen Display
A typical problem for electrical characterization in the low current range is the need to determine the achievable device performance of a low power/low leakage MOSFET under different conditions.
measurements are critical because they identify specific metrics (FoM) that confirm or deny valid behavior within a specific app. For example, n-type FETs require evaluation of the turn-on and turn-off drain currents at different values of source, drain, and gate voltages. FoMs may vary from application to application, but the way the metrics are obtained is essentially the same: provide a precisely controlled voltage or current that changes in a certain way, while simultaneously obtaining voltage and current measurements accurately and associated with each specific variable variable .
In practice, this problem can be solved by using a certain number of source measure units (SMUs), which are specialized instruments that can provide current or voltage while measuring it. But while a practical solution looks ready, there are many hidden “details” that can cause problems and misleading results, let’s take a look.
These Key Questions to Ask Yourself
Increasingly common, engineers fall into the trap of forgetting to take a closer look at the test system as a whole. Or better yet, they see their devices clearly, they see their instruments clearly, but they don't see what's in between. For example, I often see oscilloscope users forget to use probes to reach specific test points to measure the board. For those engineers who are reminded to consider the impact of probes on signals, they often still forget about the impact of probe leads on measurements and issues related to signal coupling.
"So, does it really matter?" they ask. Unfortunately, there is a relationship and we must consider these effects. The risks are similar for
and DC characterization applications. Even though we use complex and expensive detector station system components to complete physical detection, the SMUs still have to force the voltage, measure the current, and connect to the probe card through cables. Does this mean we should consider the cable as potentially affecting our measurements?
Whatever the answer is, it is important that you ask the question yourself before proceeding. More importantly, make sure your answers are correct.
CMOS Precision measurements in manufacturing are a prime example of the importance of connectivity. In fact, connectivity means adding capacitance to the test system. Since today's MOSFETs are characterized over a wide extended frequency range, any effects caused by the added capacitance must be carefully considered.
Let’s first look at the impact of connections on capacitance. Parametric (automated) test equipment is typically connected using triax cable, which is a very typical example of a low-noise connection between a source-measure test unit and the device under test. Triax is a special type of coaxial cable that insulates the signal-conducting portions with an additional outer copper Faraday shield. Even though the Faraday shield reduces the distributed capacitance of the cable, the added capacitance of the cable still affects the measurement when the total cable length becomes meaningful.
Let's look at a practical application, such as a test system that must characterize n-MOSFET transistors. In this application, we use an SMU-based test system to track the so-called I-V curve, which is sometimes called the "output characteristic" or "transfer characteristic". We programmed the gate voltage to scan forward and backward (using an SMU as mentioned before) while simultaneously measuring the drain current (also using an SMU).
Through these curves, we can collect useful data, accurately establish transistor conduction force activation and deactivation models, analyze when these characteristics reflect linearity or enter saturation behavior, and determine how the self-heating effect may affect these parameters and curves. How much displacement will occur.
When characterization requires modeling the behavior of carriers , electrons, or holes (jumping between states, modifying their mobility based on a variety of conditions), the measurement system is connected to the DUT in a four-wire (or remote sensing) configuration , and use triaxial cable.
Take a look at the triax connection in a four-wire configuration. The total length corresponds to the sum of the Force Hi and Sense Hi cable lengths. According to the capacitance/meter (pF/m) index of the triaxial cable, we can calculate that using two triaxial cables to connect the SMU to the device terminals, the length is 20 meters (10 meters + 10 meters), and the protective capacitor is approximately 2 nF, and the shield capacitance is approximately 6 nF.
In these cases, the sensitivity of the SMU is meaningless when measuring the transfer characteristics of weak currents (generally in the nanoampere range) because capacitive cable loading will cause oscillations. Not only must the SMU be sensitive, it must also be able to maintain the effective capacitance caused by cable loading, or the loading of any leads connecting the SMU to the DUT.
Otherwise, the sensitivity is useless and the SMU will just produce noisy oscillatory readings. Comparison of the Id-Vd curves of
using two SMUs and FET measured through the switch matrix using two 4211-SMUs.
Being able to determine whether the test capacitance is affecting the measurement is becoming increasingly critical. In these situations, Keithley applications engineers can provide valuable consulting services to ensure customers avoid pitfalls. When there are long connecting cables in the setup, or when there is a switch matrix between the measurement system and the DUT, or when the DUT or chuck requires nanoamp level measurements, it is a good idea to review the setup and seek advice from a consultant.
Latest solutions for critical ranges
Faced with these very challenging special situations, it is necessary to use specific SMUs modules in the measurement. Keithley has launched a special version of SMU that can be used in parameter analysis systems similar to the 4200A-SCS parameter analyzer.
SMUs are particularly suitable for connecting to LCD test stations, probes, switch matrices or any other large or complex testers. The
4201-SMU medium power SMU and the 4211-SMU high power SMU (with optional 4200-PA preamplifier ) support stable low-current measurements, even when long cable connections result in high test connection capacitance.
In fact, these modules can power and measure 1,000 times more capacitance than today's systems. For example, if the current level is 1 to 100 pA (picoamps), then the latest Keithley modules will stabilize at a load capacitance of 1 µF (microfarads). By comparison, competing models can only tolerate 1,000 pF (pico Farads) of maximum load capacitance before measurement stability deteriorates—in other words, 1,000 times worse than the Keithley module.
C-V measurements for high impedance applications
Summary
Continuous improvements in measurement technology are essential to optimize semiconductor materials to achieve low contact resistance , specialized shapes and unique structures in integrated transistors. The success of GaN transistors in future power electronics is closely related to the nanostructure used in the casting process. On the one hand, the capacitance is lower in the gate width structure, so any other meaningful capacitance effects, such as those due to cables and connections, need to be taken into account. On the other hand, they also overcome these problems by improving the SMU's ability to tolerate high capacitance, providing greater measurement stability.
