Applying Kelvin Measurements To PMIC Testing on ATE
The increasing number of power domains found in power management ICs (PMICs) targeted for portable applications places heavy demands on the few precision DC resources found in many ATE systems. Even with the recent introduction of high-density DC instruments offering both accuracy and power, many installed ATE systems don't yet have these newer instruments and, for that reason, can't benefit from the improved accuracy.
Kelvin techniques are used in many applications, including the new high-density DC instruments, to improve measurement accuracy. Testing PMICs on ATE systems offers numerous opportunities to apply Kelvin connections. Proper application of Kelvin techniques using the existing DC resources of installed systems can effectively address the number of power domains and accuracy demands of today's PMICs.
A Refresher Course
A simple resistor measurement is used to introduce the concept and illustrate the benefits of Kelvin connections. Figure 1 illustrates a two-wire resistance measurement setup including error sources. A constant current, If, is forced through the resistor under test, Rdut, and the resulting voltage is measured.
Figure 1. Basic Two-Wire Resistance Measurement
With Fixture Effects
Rw1 and Rw2 are fixture resistances from loadboard traces, pogo pins, and tester wiring. A value of 0.4 Ω to 0.7 Ω is not uncommon for typical trace geometries. The measurement error is dependent on the value of the resistor; for a 75-Ω resistor, it is about 1 to 1.9%.
Kelvin sensing eliminates this error by using four connections instead of two, resulting in the common name of a four-wire measurement. Two connections are used to force current (the force connections), and two connections are used to measure voltage (the sense connections). Figure 2 shows a Kelvin resistance measurement setup including error terms.
Figure 2. Kelvin Connection With Fixture Effects
The fixture resistances, Rw1 and Rw2, although still present, now are outside the voltmeter measurement connections and no longer contribute an error. However, two residual sources of errors exist in the Kelvin setup.
The voltage drop across Rwa and Rwb, caused by the sense current, IVM, is an obvious error. In addition, the sense current reduces the force current delivered to the DUT, producing another error. Because of the high input impedance of the voltmeter, the sense current is extremely small, which results in both of these errors becoming insignificant.
An ATE system's parametric measurement unit (PMU) acting as the Kelvin sense typically uses a current in the range of 100 nA to 10 ??A, which has a relatively small impact on accuracy for most measurements. For the 75-Ω resistor example, a sense current of 100 nA results in an error of 764 ??Ω or about 0.001%. Increasing the sense current to 10 ??A still only produces a 0.1% error. As the PMU sense current is known, the error due to reduced force current can be compensated in the test setup by increasing the force current value by an amount equal to the sense current.
Applying Kelvin to PMIC Testing
Testing high current display and LED drivers, low dropout (LDO) regulators, buck and boost switching regulators, battery chargers, and audio amplifiers found on PMIC devices is an ideal application environment for Kelvin techniques. Without Kelvin connections, the fixture resistances from loadboard traces, pogo pins, and ATE internal wiring will introduce errors. Properly and creatively applying Kelvin techniques reduces the errors and significantly improves accuracy and yield.
Many PMIC measurements either force a current and measure a voltage or force a voltage and measure a current. In either case, the current will cause a voltage drop across any fixture resistance. Typical load current ranges from tens of milliamps for drivers up to hundreds of milliamps or even amps for regulators and battery chargers. Applying Kelvin techniques will eliminate or reduce the error associated with the fixture resistance.
ATE systems provide a number of DC stimulus and measurement resources that can be applied to test PMIC devices. Some offer Kelvin sensing capabilities directly while others can be used as either a Kelvin force or sense when correctly designed into the circuit.
DUT power supplies (DPS) provide power to the DUT power pins using a four-wire connection. DPSs also are used as active loads for high-current regulator outputs as well as the power input source for the regulators. Their four-wire force and sense connections typically are dedicated to a single DUT pin or pin group although they can be shared using relays.
System PMUs (SPMUs) offer higher resolution and accuracy and increased operating range over per-pin PMUs (PPMUs). SPMUs are used for accurate voltage measurements for reference voltages and device trimming. They can be used as loads for medium-current output regulators and drivers and also to provide power input for medium and lower power regulators. The four-wire force and sense connections are brought to the loadboard and either dedicated to a DUT pin or shared across a number of DUT pins using loadboard relays.
In many ATE systems, the SPMU is multiplexed to all digital channels through an internal DC bus. Then any channel has access to an instrument with higher resolution, accuracy, and operating range. However, the four-wire Kelvin connection is reduced to a two-wire connection when used through a digital channel.
A digital channel's PPMUs and active loads are both useful in PMIC testing. Combined they provide increased current loads for regulators and drivers over the PPMU used alone. The PPMU also is useful for Kelvin sense measurements in otherwise non-Kelvin circuits with a little upfront planning and design.
Example PMC Kelvin Measurements
A 1.8-V LDO output capable of supplying 180-mA current is tested for voltage accuracy and load regulation. Without Kelvin sensing, what is the error due to the uncompensated series resistance of the contactor, loadboard trace, pogo pin, and internal resistance to the SPMU?
The series resistance may be in the range of 0.2??Ω to greater than 1??Ω depending on the loadboard layout. For this example, assume the pogo pin and SPMU internal path is 0.2??Ω, the loadboard 0.2??Ω, and the contactor to DUT pin interface 0.1??Ω series resistance for a total of 0.5??Ω. Without a Kelvin sense connection, the error is:
Verr = 180 mA x 0.5 Ω = 90 mV
This is about a 5% error, which is equal to or greater than a typical accuracy specification and well over a typical load regulation specification.
Proper Kelvin sensing can be configured in several ways. The simplest solution is to use either a DPS or SPMU with their direct Kelvin connections dedicated to the DUT pin.
The +Force, +Sense, -Sense, and -Force terminals are connected to the corresponding +Frc, +Sns, -Sns, and -Frc terminals. For the SPMU, the -Force connection often is the tester ground reference and need not be connected on the loadboard. The DPS and SPMU sense line currents are less than 10 nA, so the error introduced is negligible.
The only residual error is caused by the contactor resistance. In this example, the error is 18 mV or 1% which is now within accuracy specifications.
Another method routes the SPMU through a digital channel using the system internal DC bus. This approach is useful if several DUT pins require a current load greater than the combined capacity of the PPMU and active load. The SPMU +Force and +Sense are connected internally in this mode, and -Sense is internally referenced to tester ground. These connections are shown in Figure 3.
Figure 3. Kelvin Connection Options Using SPMU or PPMU
A Kelvin sense still can be realized by using another digital channel's PPMU as the sense connection. The current used in the PPMU sense line is still very small, typically 100 nA, so its effect will be negligible. Again, the remaining error is due to the uncompensated contactor resistance.
The same Kelvin sensing technique can be applied using only digital channel resources. Multiple digital-channel PPMUs combined with active loads can be paralleled to increase the current capability. An extra digital channel then provides the Kelvin sense connection. With this approach, multiple measurements can be performed in parallel for shorter overall test time.
The error due to the contactor resistance can be eliminated if the +Sns connection is made directly at the DUT pin. Some DUTs use multiple pins for high-current regulator outputs. Using one of the pins as a Kelvin sense connection, the residual error due to the contactor resistance is eliminated. The remaining pins must safely handle the full current load.
In some applications, a resistor on the loadboard is effective as a current load. Using a resistor load saves valuable high current tester resources. Loads for audio amplifiers, found in some PMICs, and for LDO regulators are good candidates for this application.
This application is very similar to the resistor example used to introduce Kelvin measurements. An audio amplifier will illustrate the difference in the design approach required to correctly measure according to device specification. Usually, the voltage at the DUT pins driving into a specified load is required.
A resistor on the loadboard simulates the speaker load. A high-input impedance digitizer measures the signal. The circuit in Figure 4 shows the amplifier, load resistor, digitizer, and simplified loadboard trace resistances for two connection options.
Figure 4. Kelvin Connection Using Load Resistor for AC Measurements
The B+ and B- connection may appear to be a proper Kelvin connection; however, in this case, the A+ and A- connection is the correct choice because the amplifier output voltage at the DUT pins is specified. If the connection is made to B+ and B-, the voltage drop across Rw1 will reduce the measured value. Although trace resistance Rw1 can be minimized by keeping Rload as close to the amplifier as possible and using heavy traces, it is not possible to completely eliminate it.
The A+ and A- connection measures the voltage at the amplifier pins as specified. An error still exists due to the series resistance Rw1, but if the value of Rload is reduced by 2Rw1, this error also can be eliminated.
Assuming the Rload value is reduced, the A+ and A- connection creates a correct measurement setup and compensates for the residual error. It is not possible to compensate for the error when using the B+ and B- connection.
During loadboard layout, the designer may choose the B+ and B- connection because of the ease of physical layout. For that reason, it is important to clearly convey the proper layout requirements.
The digitizer's 1-MΩ input impedance minimizes the effect of Rw2 on measurement accuracy. When a 50-Ω input impedance digitizer is used, it typically acts as the load. In this case, the effect of Rw2 needs to be considered.
The concept similar to that illustrated in Figure 4 applies when using loadboard resistors for LDO DC measurements. Substitute the DUT's DC output for the amplifier and a PMU for the digitizer. The same considerations apply when dealing with trace resistances and layout.
Guideline for Kelvin Connections
There are many rules of thumb about when to use Kelvin connections including when the current is greater then 100 mA or 500 mA and when the resistance is below 10 Ω or 100 Ω. These are all useful in particular situations. A pragmatic guideline for when to use Kelvin connections is when not using them has a measurable impact on the overall measurement accuracy or yield.
Of course, this requires an upfront analysis of potential error sources, calculating error magnitudes, and comparing the results to the total error budget for yield impact for each measurement. I've established the following practical and customized guideline that is easy to apply: Consider using Kelvin connections when not using them would increase the error budget by 5% to 15%.
Conclusions
As the examples have shown, Kelvin connections provide benefits even at relatively low currents. By using Kelvin techniques for PMIC testing, you can make the best possible use of your existing ATE resources to effectively test today's PMICs.
Applying careful layout and design practices will allow you to understand when using a Kelvin connection will be beneficial. The extra analysis and design required to implement Kelvin connections can significantly improve accuracy and yield.
About the Author
Kent Luehman is a lead technical consultant for mixed-signal SOC test in the Techinical Expertise Center at Verigy. He has more than 32 years experience in application engineering, R&D, and marketing for ATE, including time at Agilent Technologies and Hewlett-Packard. Mr. Luehman began his career at the H-P New Jersey Division where he was on the design team for the first HP-IB controlled power supplies, earning three patents. Verigy, 10100 N. Tantau, Cupertino, CA 95014, 408-864-2900, e-mail: [email protected]