Observing wafer-level device behavior via measurements in the femtoamp range—a procedure prompted by today’s submicron device technology—is tricky business. More specifically, measuring current values in the 1-fA range is virtually impossible to achieve without special probing equipment.
Even with such equipment it’s very challenging. The quantity of current is so low that a minuscule flow of electrons generated from other sources can interfere with the measurement. To put this in perspective, consider this: 1 fA (1 X 10-15 A) is equal to a flow rate of 6,242 individual electrons per second compared to the 6.242 X 1018 electrons per second that defines an ampere.
Low-Current Measurements
Grounding and shielding techniques required for measuring low current use two typical probe-contact configurations: surface-to-surface and surface-to-substrate. For this discussion, the surface-to-surface configuration is represented by MOSFET subthreshold measurements and the surface-to-substrate by time-dependent dielectric leakage (TDDL) measurements (Figure 1).
Subthreshold measurements permit the analysis of drain-current behavior below the gate turn-on voltage. This off-state drain current, although less than 100 fA, will require higher DRAM refresh rates.
TDDL characterization identifies the behavior of gate-oxide leakage under the influence of high electric field stress to thin insulating layers, such as oxide. This behavior can affect the reliability of EPROM and flash memory devices because write/erase cycles will cause an increase in leakage current over time. This current will be in the femtoampere range and can increase component leakage and affect other aspects of device performance, resulting in eventual circuit failure.
Source Measure Units
Subthreshold and TDDL characteristics are graphically analyzed and both require a voltage for the stimulus or source, which produces a current as the response or measured value. A source measure unit (SMU) is commonly used to facilitate this stimulus and response function. An SMU can be configured to source current and measure voltage or to source voltage and measure current.
SMUs have two features that facilitate measurement accuracy. When IR drops in the cabling between the SMU and the DUT are a concern, remote voltage sensing is available. If leakage paths and stray capacitive charging in the fixturing approach expected DUT current levels, a guard voltage output can minimize the influence of these paths. Figure 2 shows the configurations that will be used in this article.
Factors Affecting Low-Level Environment
Many factors can adversely affect low-level measurements. Use the following descriptions and suggested preventive measures to visualize the relationships between noise and signal and to alter your experimental measurement setup to better isolate the noise from the signal. Remember that noise-coupling mechanisms are best described as “virtual components” because they will not be obvious or localized as would a lumped equivalent component.
Common Impedances: A common impedance is an impedance or current path unintentionally shared by a source of noise and the receiving instrumentation. The noise source causes current to flow through the common impedance.
If this common impedance is in parallel with the instrument’s input, a noise voltage will be coupled into the measured signal with the relationship:
Vnoise = inoise (Zcoupling)
If the instrument is measuring current, then the current flowing through the common impedance will algebraically sum with the signal current at the connecting node.
Prevention Measure: Analyze the grounding scheme to verify that it conforms to the single-point grounding practice. Confirm that current paths between different circuit functions are isolated. Inspect the physical test setup and eliminate all casual, unintended contact to or between ground conductors that reside in different electrical parts of the circuit.
Incidental Mutual Inductance: Transfer of a noise voltage can occur through mutually coupled inductances in the noise circuit and the signal circuit. This inductive coupling will have the relationship:
Vnoise = MWhere M is the mutual coupling between the noise circuit and signal circuit and will be influenced by proximity and shared length of the noise and signal conductors.
Prevention Measure
: Eliminate common runs of conductors and cables; especially conductors not part of the test signal circuitry. Maintain wide conductor spacing where possible. Isolate circuits with magnetic shielding where practical.
Magnetically Coupled Noise: Typical noise sources that lend themselves to magnetic coupling are motors, transformers and facility wiring. The coupling mechanism can be expressed as:
Vnoise = 2p ¦ BA cos q
Where: f is the frequency of the noise signal
B
is a measure of the noise magnetic field strengthA
is the area of a closed loop in the signal circuit
q is the angle of B to the plane of the loop area A
Prevention Measure: Remove unnecessary sources of electromagnetic energy in the area of the test circuit. Eliminate circuit loops. Magnetically shield wires or use twisted pairs.
Incidental Capacitive Coupling: A source of noise current can be coupled to the measurement circuit via stray capacitance according to the relationship:
inoise = CcouplingWhere Ccoupling is the stray capacitance between the noise source and the measurement circuit.
Stray capacitance is inversely proportional to the distance between conductors. If the circuit of interest has a high impedance, it will see a noise voltage. If the impedance is low, a current will be developed.
Prevention Measure: Add distance between unshielded noise sources and the test circuit. Separating the two wires by a distance greater than 40 diameters provides good electrostatic isolation.
Ensure the test circuit conductors are shielded to direct capacitively coupled current away from the test circuit to ground. Active guard shields also provide good protection.
Charge Transfer: If a charged body is brought in proximity to the test circuit, that charge will attract or repel the charge in the test circuit and cause momentary electron flow. This perturbs femptoampere measurements. A human body is the typical charge vehicle.
Prevention Measure: Limit activity in the area of the test setup during measurements. Provide broad electrostatic isolation by using a conductive light-tight enclosure (LTE). Run interconnecting cables between the LTE and the SMU behind the test bench and away from traffic and AC power cords.
Stray Capacitance Charging: The voltage supplied by the SMU for subthreshold and TDDL measurements is typically a stepped sweep of increasing voltage over time. Stray capacitance in the cables and shield surfaces between the SMU and DUT must be charged, which introduces an added current that swamps the DUT current. For low-current measurements, this is a major source of error.
Prevention Measure: Use an active guarded shield on all cables to keep the differential charging voltage to a very small level. Use probes or probe holders with guard shielding as close as practical to the probe point. Evaluate the settling-time requirement after the guarding is in place, and be sure to program an adequate measure delay time into the SMU.
Light: Semiconductor devices can be light sensitive. This is especially true for MOSFET devices that rely on minority carriers for conduction. At femtoampere levels, the currents induced by exposure to light overwhelms many measurements.
Prevention Measure: Use a conductive LTE to provide a dark, shielded environment around the DUT. This blocks all forms of electromagnetic radiation, not just the visible spectrum.
Intrinsic Noise Sources: At low femtoampere levels, most coaxial and triaxial cables generate current by triboelectric effects caused by flexing or vibration of the cable. Contacts between dissimilar metals, as might be found in connectors, crimp joints or solder joints, develop temperature-sensitive Seebeck voltages (thermal noise). Moisture combined with chemical residue left over from an assembly process produce an electrochemical voltage.
Prevention Measure
: Minimize triboelectric errors by allowing cables to relax after installation. Anchor cables to prevent motion. Prevent Seebeck effect and electrochemical voltages by making connections with similar metals and keeping connectors clean and dry.
Leakage Current: All insulation used in cabling, connectors and probes permits a degree of electrical leakage that can influence femptoampere measurements. Some of this leakage is inherent to the insulating material, and additional leakage may sometimes be contributed by contamination during assembly or handling.
Prevention Measure: Use an active guarded shield on all cables to keep the differential voltage between the signal conductor and the shield to a very small level.
Successful Low-Level Measurements
The following procedures and equipment considerations will aid in achieving success in a low-current probing application.
Diagnose Noise Sources; Minimize Their Effects: Solving noise problems at femtoampere levels is difficult because analog time-domain measurement techniques that permit direct observation of the low-level noise are not readily available. When a noise waveshape can be seen in real time, its behavior can be compared to external events and coupling mechanisms can be deduced and modified to reduce their influence.
Since this luxury probably will not exist, the next best tool will be your parametric tester configured to provide a sampled sweep of current vs time with a constant source voltage. When used this way, the filtering and averaging options should be set to a minimum to improve the tester’s use as a noise diagnostic tool. By observing the sampled noise in this way, possible sources and coupling paths can be identified and altered while observing the effects of the alteration.
Check System Grounds, Shields, Connections
: The ideal grounding, shielding and connection scheme for surface-to-surface and surface-to-substrate current measurements is shown in Figures 3 and 4. Note the use of single-point grounding and active guarding.
Exploit Available Filtering, Averaging: Parametic measurements like sub-threshold and TDDL lend themselves to filtering and averaging. Use these features to obtain a clean, nominal representation of these DC characteristics.
Use a Guarded Signal Path: This will be a key contributor to successful low-current measurements. The driven guard output on a typical SMU will provide a buffered output voltage that follows the source output. Since the voltage difference between the probe and the cable signal path and the guard shield will be nearly zero, the leakage current through the isolation will be maintained at very low levels.
Charging delay (settling time) of the signal line will be substantially reduced for the same reason. Refer to Figures 3 and 4 for the recommended guard configuration. Compensate for the remaining settling time by adding the appropriate measurement delay to the SMU.: Femtoampere current measurements require very high resistive isolation and guarding between the signal path and all conductors at reference potential. Most cabling generates some degree of intrinsic noise, especially when the cable is flexed or vibrated, but most of this will subside after the motion stops.Assure Cable Isolation and Low Noise
Cable designated as low noise is manufactured to minimize triboelectric effects. When evaluating cables for leakage and intrinsic noise, allow the cable to relax into a limp position before making a leakage measurement. The noise-settling time could vary from seconds to tens of minutes depending on the cable type and characteristics.
Maintain Cleanliness: Avoid touching insulation between two conductors such as might be found around connectors, probes, shields and guards. Oils or contamination on your fingers could provide an impedance path and cause an increase in leakage current.
Hardware Considerations
Probing Station: The probing station best suited for low-current measurements has a chassis configured with a single-point grounding scheme to minimize current generated by Seebeck voltages in the chassis. The station should also feature isolated, independent electrical access to the chuck surface and chuck guard when used with a triaxial chuck (Figure 4).
Low-Noise Enclosure: The enclosure should provide a dark environment and electromagnetic shielding from external sources of noise. A completely enclosed probe station offers convenient access to the wafer, probe and probe holder. An integrated environment can offer the same protections but can be really difficult to set up and use.
Take special care when measuring at temperature extremes. Such measures can be satisfied via a special environmental enclosure that prevents moisture and frost formation on the DUT.
Low-Noise Chuck: A triaxial chuck, shown schematically in Figures 3 and 4, provides very high isolation to ground (with or without guarding).
Probe Holders: Specially designed probes (holders and needles) are available for low-current probing. Use a probe with triaxial cabling, which permits the use of the inner shield as a guard within close proximity to the replaceable probe tip.
For applications where both high and low currents are present, a Kelvin-type triaxial probe holder should be used to compensate for IR drop in the source voltage cable.
Guarded probing is possible using a coaxial probe holder in conjunction with an adapter that routes the inner shield from the triaxial SMU cabling to the probe shield so that the shield continues the guard down to the vicinity of the probe tip. With this type of probe holder, the guard voltage will be accessible at the BNC connector body. Consequently, a potential shock hazard exists with the use of these probe holders, especially if the guard voltage exceeds 30 V. The connector should be isolated and insulated and handled only when the guard voltage is off. When using this type of probe, use an external LTE to shield the probe cable and station.
Cables and Adapters: Use triaxial cables intended for low-noise applications for low-current probing. Specify low-noise, low-triboelectric-effect cabling.
When adapters are required to connect BNC connectors to triaxial connectors, use ones that connect the triax inner shield to the BNC shield.
Summary
The need exists for low-current measurements in the 1-fA range due to the continuing design trend toward decreasing geometric scale in devices. With this reduction in size, however, finding the desired low-frequency signals associated with such geometry amid the noise poses many challenges.
The factors that affect measuring in low-current environments can be modeled as virtual components. These models help you visualize the noise sources and noise-coupling mechanisms when trying to trace and eliminate error sources in the test system.
Accurate low-level measurements require a low-noise, low-leakage and low-capacitance probing environment. The best way to provide such an environment is to use tools and techniques that allow you to prevent or negate the sources of error.
All of the prevention measures and guidelines presented in this article have been proven at The Micromanipulator Company using a low-current probing instrument package. Reference 1 contains equipment configurations and actual test measurements.
References
1. Application Note A1009492, “Basics of Low Current Probing”, revision B, The Micromanipulator Company, Inc., April 1995.
Acknowledgment
The author credits Jim Forst, formerly of The Micromanipulator Company and now of Methodics, for providing the basic structure for this article.
About the Author
Michael S. Jackson, who has been affiliated with analytical probing since 1988, is the Director of Marketing and Sales at The Micromanipulator Company. Previously, he held R&D, production management and engineering, and applications management positions in the semiconductor and LCD industries. Mr. Jackson is a graduate of Regis College with a B.S. degree in technical management. The Micromanipulator Company, 1555 Forrest Way, Carson City, NV 89706, (702) 882-2400.
Copyright 1996 Nelson Publishing Inc.
March 1996