Though the error is small, the idea of canceling the effects of unbalanced input bias currents by purposefully misbalancing the input traces is neither practical nor recommended. The difference in input bias currents will vary part-to-part, wafer-to-wafer, and lot-to-lot. It can also vary based on design parameters such as common-mode voltage and supply voltage.
Therefore, it is recommended to place the current sensing device and shunt resistor on the same side of the PCB board and to use balanced, short input traces.
As shown in Equation 6, parasitic trace resistance in series with the shunt resistor can induce measurement error. In this scenario, the parasitic resistance is in the high current path. Therefore, it does not take a significant amount of parasitic trace resistance to cause significant error.
For example, a one-ounce copper trace of 10mils in length and 25mils wide has an impedance of approximately 193µΩ. While this may not seem significant, please recall that it can carry the maximum load current (10A). Table 4 summarizes the effects of such input traces on the circuit from Figure 3 with a load current of 10A.
Using the ideal shunt voltage value of 80mV, we calculate the error due to Rps using Equations 1 and 2 as follows:
Equation 9 depicts the importance of removing all parasitic resistance in series with the shunt resistor. This can be accomplished by implementing a Kelvin-connection between the shunt resistor and the current sensing device, which will be discussed in the section on Layout Recommendations.
While this error is clearly more significant than the errors caused by Rpp and Rpn, an additional observation can be made: it is possible to measure a larger voltage at the input pins of a current sensing device than across the shunt resistor. Consequently, a larger voltage will be observed at the output of the device.
This observation is important when troubleshooting a current sensing solution. It is common for one to measure the voltage across the shunt resistor, multiply it by the nominal gain of the device, and compare the result to the measured output of the device. In this case, the ideal shunt voltage is 80mV, the nominal gain of the device is 50V/V, and the simulated output is 4.084V. The product of the shunt voltage and nominal gain is 4V. The 84mV discrepancy (or 2.1% error) could lead one to believe the device’s gain error is not within specification.
The error introduced by the device itself, however, can be calculated using Equations 10 and 11:
The use of V’sense for this calculation eliminates any error induced by the PCB layout and reports only the error introduced by the device itself.
We see the error contributed by the device is actually 0.3% when V’sense is used for the calculation, which is significantly less than when using Vshunt. When troubleshooting a current sensing solution, it is imperative to measure the voltage at the input pins of the device instead of the shunt voltage.
It has been shown that any parasitic resistance in series with the shunt resistor will induce an error. Therefore, it is recommended that the designer implement a Kelvin-connection between the shunt resistor and the current sensing device. Non-Kelvin and Kelvin connected layouts are shown in Figures 4, 5, respectively.
As discussed earlier, it is recommended that the input traces are short and balanced. Figure 6 depicts a layout such that Rpp>Rpn.
We showed that parasitic resistances either in series with the shunt resistor or in series with the input pins of a current sensing device can induce measurement errors. In order to minimize such errors, it is recommended to Kelvin-connect the shunt resistor and use short, balanced traces for the input connections. Finally, care should be taken when calculating device error. It is recommended to take all device-related measurements at the pins of the device as opposed to other convenient locations, such as test points or across the shunt resistor itself.
1. Karki, James. “Understanding Operational Amplifier Specifications,” White Paper (SLOA011), Texas Instruments, 1998.
For answers to current sensing applications questions, visit TI’s Precision Amplifiers forum in the E2E community.
Download a free copy of TINA-TI here.
About the Authors
Peter Semig is an Applications Engineer in the Precision Linear group at TI where he supports difference amplifiers, instrumentation amplifiers, and current-shunt monitors. Peter received his BSEE and MSEE from Michigan State University, East Lansing, Michigan. If you have questions about this article, contact Peter at firstname.lastname@example.org.
Collin Wells is an Applications Engineer in the Precision Linear group at TI where he supports industrial products and applications. Collin received his BSEE from the University of Texas, Dallas.