Content area
Full Text
As more systems become electrified, thermal management has turned into one of the "hottest" issues facing designers. Using current measurements for thermal management is a leading indicator of system performance and faults, whereas simply monitoring the temperature is potentially a lagging indicator. Accurately monitoring the current consumed, especially over temperature, has become vital as designers pack more functionality into tighter areas.
While room-temperature calibration tends to be relatively straightforward, performing multi-temperature calibration is time-consuming and costly. Identifying ways to minimize the effects of temperature on current measurements can improve system performance and minimize system design margins, as well as potentially lower the total cost of ownership (TCO).
SOURCES OF ERROR IN CURRENT MEASUREMENTS
As I stated in my September 2015 article, "Mitigate Error Sources to Maximize Current-Measurement Accuracy" (see http://www.electronicdesign.com/test-measurement/mitigate-error-sources-maximize-current-measurement-accuracy ), there are multiple contributing sources of error in current-measurement applications. In the article, I listed these sources of errors:
Amplifier-related errors:
Input offset voltage (VOS) and VOS drift
Common-mode rejection ratio (CMRR)
Power-supply rejection ratio (PSRR)
Gain error and gain drift
System errors:
Gain-setting network tolerance, matching, and drift
Printed-circuit-board (PCB) layout
Shunt-resistor tolerance and drift
You can see that "drift" is part of four of the seven items on both lists, which emphasizes the importance of minimizing the additional errors caused by temperature in a current-measurement implementation.
DISCRETE CURRENT-MEASUREMENT IMPLEMENTATIONS
Many system designers choose a discrete amplifier and external gain network for their low-side current-measurement applications because it's viewed as a low-cost alternative. There are two options when using a discrete circuit for low-side current sensing: a single-ended or differential configuration. Figure 1 shows the latter.
In either configuration, the gain of the system is defined as G = RF /RI . The worst-case initial (or room-temperature) gain error is simply the tolerance of the discrete gain resistors. Assuming an application with a gain of 20, where RF = 100 kΩ and RI = 5 kΩ, Table 1 reveals how that looks for...