Publishing Venue
Motorola
Related People
Authors:
Abstract
Wafer-level capacitance measurement results often contain significant amounts of error. Even when proper measurement system setup has been performed, differences in measured values between test systems on the order of a few hundred femto-Farads are commonly seen. Proper measurement system setup in this context includes proper frequency, network model selection, bias, and pin connections, and subtraction of parasitic test system capacitance. This document identifies the primary reason for these differences as a Z-dependent parasitic test system capacitance. Specifically, system capacitance measured with probes separated from the wafer is not the same as system capacitance with probes in contact. The Z-dependence of system capacitance is also a strong function of probe card. The MZ capacitance measurement method is presented, which takes system capacitance measurements at different wafer to probe separations and extrapolates to determine the system capacitance at contact. Using this method, measurement offsets between test systems can be reduced significantly.
The MZ
Capacitance Measurement Method:
A Method of
Achieving High Accuracy Direct Wafer-level
Capacitance
Measurements by Taking Multiple
Test System Capacitance Measurements at Different Wafer to Probe
Separations
Andrew P. Hoover
Abstract
Wafer-level capacitance measurement results often
contain significant amounts of error.
Even when proper measurement system setup has been performed,
differences in measured values between test systems on the order of a few hundred
femto-Farads are commonly seen.
“Proper measurement system setup” in this context includes proper
frequency, network model selection, bias, and pin connections, and subtraction
of parasitic test system capacitance.
This document identifies the primary reason for these differences as a
Z-dependent parasitic test system capacitance. Specifically, system capacitance measured with probes
separated from the wafer is not the same as system capacitance with probes in
contact. The Z-dependence of
system capacitance is also a strong function of probe card. The MZ capacitance measurement method
is presented, which takes system capacitance measurements at different wafer to
probe separations and extrapolates to determine the system capacitance at
contact. Using this method,
measurement offsets between test systems can be reduced significantly.
Problem Description
Accurate capacitance measurement is required for
process monitoring and for circuit simulation model creation. Making accurate capacitance
measurements, however, is a challenging task, especially when making
measurements on an automated test system with a probe card. There are many aspects that must be
properly considered when configuring the test system. Bias voltage and frequency must be properly set, appropriate
pin connections must be made, an appropriate network model must be selected to
extract capacitance from complex impedance, and parasitic test system
capacitance must be subtracted from the measurement.
Traditional “2Z” capacitance measurements (Fig. 2a)
attempt to measure system capacitance by separating the wafer from the probe
tips by a small distance (~10 mils), and measuring capacitance. Then, the wafer is put in contact with
the probe tips, and a contacted measurement is made. By subtracting the contacted measurement from the system
capacitance measurement, the capacitance of the device under test (DUT) is
obtained.
The “2Z” capacitance measurement method can result
in measurement differences between test systems on the order of a few hundred
femto-Farads for the exact same DUT.
This is despite the fact that common capacitance meters, such as the
HP4284, have absolute accuracy which is much better than this. The reason for the measurement
differences lies primarily in inaccuracy of system capacitance
measurement. The “2Z” method
ignores the fact that system capacitance is a function of wafer to probe
separation. This Z-dependence has
been found to be the primary cause of observed system-to-system offsets.
The Z-dep...