Comparison Of The Feedback Charge Method With The Traditional Ramp Method - Keithley 595 Instruction Manual

APPLICATIONS
5.3.6 Comparison of the Feedback Charge
Method with the Traditional Ramp Method
The traditional ramp method of measurifig quasistatic
capacitance originally proposed by Kuhn, Castagne, and
Kerr (see Bibliography, paragraph 5.4) has been used ex-
tensively for MOS characterization. In the ramp method,
shown in Figure 5-15, the device under test (DUT) is con-
nected between a linear ramp generator, dV/dt, and a feed-
back current~amplifier. The current flowing through the
DUT is thus measured by observingv.., If the DLJT were
an ideal capacitor, this current would be the capacitor
displacement currentz I = C=(dV/dt).
For a capacitance of 1OOpF and a ramp rate of O.lV/sec, a
current of 1OpA would be measured. Of course, for an&z-
tual MOS capacitor, any additional current due to leakage
or non-equilibrium behavior would also be measured. The
inability to distinguish between displacement current and
other currents is a major limitation of the ramp method
when the DUT or the system has significant~leakage or
when the ramp rate is very low.
The rainp method has several other limitations. The circuit
is that of a different&or. This means that any noise appear-
ing across the DUT will be amplified by a gain of (27fC&).
The frequency dependence of this gain makes the measure-
ment particularly susceptible to interference that has high
frequency components. A graph of gain vs. frequency is
shown in Figure 5-16. Typical signal frequencies with thii
circuit are very low (below lHz). At these low frequencies
the gain of the ramp method circuit decreases, further
reducing the signal to noise ratio of the measurement. Also,
feedback resistors needed to amplify low currents must be
extremely large, perhaps lOOGo or more. Resistors this large
have poor temperature &ability, high Johpson voltage noise,
and long time constants when combined with the capaci-
tance necessary to stabilize the circuit. Accuracy of the ramp
method is limited by these and other effects to approxi-
mately 1 to 5% at low ramp rates.
The measu?ement method used by the Model 595 avoids
many of the limitations of the ramp method and thereby
achieves a more accurate measurement of quasistatic
capacitance. Figure 5116 shows some of the feedback charge
method properties. With this method, a signal charge is
measured, instead of a signal current as in the ramp
method. Currents flowing into the integrator in addition
to the signal charge can be measured independently us-
ing Q/t. The capacitance measurement can be corrected for
the effect of error currefits when it has been determined
that the current is not due to non-equilibrium in the DUT
To correct for lea&age with the ramp method, multiple
curves would have to be taken using different ramp rates.
If the conditions under which sequential curves are taken
are sufficiently repeatable, then the DC current-could be
separated from the true signal cm-rent by solving
simultaneous equations.
In the frequency range'of interest, the circuit of the feed-
back charge method is not frequency-dependent. This is
illustrated in Figure 5-16. The ratio of signal gain to noise
gain for the circuit&es not decrease when long delay times
are used. Long delay times in the feedback charge method
Et;o;yhly analogous to low ramp rates in the ramps
5-17