Improve Current Measurement Accuracy by Skewing the Input Offset Voltage on Current-Sense Amplifiers
IntroductionCurrent-sense amplifiers are sophisticated ICs, popular in electronic equipment that monitors load currents in real time. System controllers use this load information to implement power-management algorithms that modify the load-current characteristic itself, and to implement flexible overcurrent protection schemes.
Current-sense amplifiers magnify a small differential voltage while rejecting the input common-mode voltage. In this role they operate like traditional op amp-based differential amplifiers. There is, however, an important difference between these two amplifier architectures. The input common-mode voltage for current-sense amplifiers is allowed to exceed the power-supply (VCC) voltage. When, for example, the MAX4080 current-sense amplifier is powered from VCC = 5V, it can still withstand an input common-mode voltage of 76V. By using unique amplifier architectures, current-sense amplifiers are not hampered by the common-mode rejection limitations (CMRR) that arise from mismatched resistors. The MAX4080, for example, has 100dB (min) DC CMRR. In contrast, the performance of traditional op amp-based differential amplifiers is negatively impacted by CMRR, and their effective input VOS is magnified through the signal chain.
Figure 1. The MAX4080 is a precision unidirectional current-sense amplifier.
Using Calibration to Improve PrecisionThe MAX4080 has a precision ±0.6mV (max) input offset voltage (VOS) at 25°C, and ±1.2mV (max) VOS over the -40°C to +125°C temperature range. Some applications, however, need to further calibrate input VOS to enhance precision of the final current measurement. To do this calibration, VOS is typically measured during production and stored in firmware. This VOS is then adjusted digitally in real time when the equipment is in the field and in use.
A preferred method of calibration, intended for the convenience of manufacturing, would measure the VOS when there is zero load current (zero input differential voltage). In this approach one could measure the output VOS and subtract this voltage from all future measurements. Unfortunately, this method has a drawback. The VOL (output voltage low) and input VOS specifications interact, causing the output voltage not to reflect the input VOS accurately. This interaction is, in fact, characteristic of all single-supply amplifiers.
Consider the example of the MAX4080T with a gain of 20, and hypothetical zero input VOS. One would then expect to measure a true zero at the amplifier's output. However, even at zero input differential voltage, the amplifier is not guaranteed to output a voltage below 15mV (with 10µA sink current). In fact, if the output voltage measurements were used directly for VOS calibration, the amplifier would appear to have a 0.75mV input VOS (15mV/20 = 0.75mV).
Similarly, if a MAX4080T did have VOL = 0, then a positive input VOS will produce a positive output VOS as expected. However, a negative input VOS will not be "reflected" in the output voltage measurement, since the amplifier cannot really output a voltage below ground. Thus to summarize, one cannot "directly" use an output voltage measurement with zero input differential voltage to calibrate the input VOS.
There are two methods typically used to calibrate VOS during production:
- A bidirectional current-sense amplifier like the MAX4081 is used with a reference voltage of about 1.5V. This effectively translates the output voltage measurements by 1.5V, so a zero input differential voltage will output 1.5V ±VOS-induced errors. Since this 1.5V voltage is well above the amplifier's VOL, it does not affect error analysis. VOS errors can thus be calculated by measuring the difference between the output voltage and the ideal 1.5V input reference voltage. There is a drawback to this method: reduced dynamic range. The 0V–5V input range of the ADC is now reduced by 30% to 1.5V–5V. Additionally, the method requires a more expensive bidirectional current-sense amplifier to be used for unidirectional measurements. Finally, generating a low-drift 1.5V reference voltage or spending a second channel to measure this 1.5V reference voltage is not attractive.
- A two-point measurement method is employed in which two known values of a differential input voltage (load currents) are applied to the current-sense amplifier. First, straight-line approximation is applied to the output voltage measurements to calculate the input VOS by extrapolating to a zero-sense voltage. Secondly, that voltage measurement is then used for calibration. This method has its drawback: it uses two "known" exact values of currents during production, which is inconvenient and increases test time. Finally, accurate measurements close to a zero input differential voltage are still not achieved because VOL limitations will cause errors at small sense voltages.
Using Input Resistors to Introduce Input VOSThis application note presents a third method for measuring the input VOS of a current-sense amplifier. Once again, the MAX4080 serves as the example. This approach applies a zero input differential voltage and overcomes the interactions between VOL and VOS—making it easy to use on the production line.
All current-sense amplifiers have input bias currents. Therefore the use of input resistors (for input filtering, as an example) should be carefully studied since the resistors can introduce unplanned gain and offset errors. These issues are detailed in application note 3888, "Performance of Current-sense Amplifiers with Input Sense Resistors." The method presented here uses similar techniques, but in this case, the input resistors are intentionally mismatched. In this manner an intentional output VOS is introduced. The MAX4080 has a temperature-compensated bias current of 5µA (typ) and 12µA (max) over process variation. Using a 2kΩ resistor in series with RS- (Figure 2) thus produces a typical input VOS of 10mV and 24mV, respectively, over process variation. This additional input VOS then causes an output offset of 200mV (typ) and 480mV (max), which is adequate to override any VOL and VOS limitations in the basic MAX4080. Error in this input-resistor-induced VOS will have a temperature dependence based both on the drift characteristics of the input resistor (usually 100ppm) and on the bias current (negligible).
Figure 2. The MAX4080 configured to use an external 2kΩ resistor in series with RS-.
The resistor drift characteristic of +100ppm causes a +1% change in the resistance value over a 100°C change (i.e., +20Ω). Additional input VOS drift from the input resistor is then typically about +0.1mV, and +0.24mV max across the process variation of bias current. This drift is still only 20% of the ±0.6mV bidirectional error in input VOS that one would usually expect from process variation if no calibration were used.
To account for the 15mV VOL and the ±1.2mV input VOS over temperature, the additional input VOS would need to be a minimum of
Table 1. Results of Temperature Tests With and Without Input Resistors
|No Input Resistors||-0.015mV||0mV||-0.005mV||-0.01mV|
|2kΩ in Series with RS-||9.69mV||9.73mV||9.76mV||9.80mV|