Many of today's applications require current measurement for power estimation, regulation purposes, or for diagnostics. This article focuses on circuits with analog outputs that are needed for loop regulation or for signal conditioning for ADCs.
In general there are two ways to measure currents – high side or low side, as shown in Figures 1 and 2. Low side current measurement seems to be the easier choice since the signal chain, in this case, is ground related. In practice, however, it might be impractical if the application circuitry cannot be raised from ground because ground integrity has to be obtained.
Both methods have one thing in common – a voltage difference, which is measured across a precision resistor or, a shunt resistor. Typically, these resistors have very low resistance (in the lower mOhms range), high accuracy and low thermal drift. It is important to know these figures in order to make an error budget calculation. A shunt is specified by its tolerance, for example, +/-1%, and the TCR (thermal coefficient of resistane) in ppm (parts per million). 20ppm/°C can be regarded as a very good shunt, while 500ppm/°C would be deemed average.
The following calculations give the TCR errors:
T = (85°C – 25°C) = 60°C
TCR Error (20ppm shunt) = 60°C * 20ppm/°C * (10exp-4)% = 0.12%
TCR Error (500ppm shunt) = 60°C * 500ppm/°C * (10exp-4)% = 3%
Because the shunt itself is expensive, the more accuracy and low drift that is required, so a compromise between the tolerance and TCR has to be found. In addition, errors in other parts of the signal conditioning circuitry need to be as low as possible. Both parameters form part of the error budget along with the offset and offset drift of the active components used and the errors caused by the resistors, which are often overlooked parts in the chain.
Error = Shunt + TCR + Offset OP + Drift OP + Gain Error R + CM Error R
It is commonly required that the shunt be made small so as to reduce power losses. However, the voltage drop has to be sufficient to reduce the amount of gain needed in the later stages, since all shunt and active component errors – such as offset and temperature drift – will be amplified. Table 1 shows typical currents of 5A – 50A and the corresponding shunt resistor values necessary to keep the maximum power losses to around 1W. Keeping the shunt size large can become more difficult depending on the heat generated by the shunt resistor and also on its temperature coefficient, which can lead to error due to thermal drift. In this case, the shunt size can be reduced but this leads to lower full scale shunt voltages, as shown in Table 2. Select the shunt as small as necessary, but as large as possible!
Table1: Shunt power losses (W)
Table 2: Shunt full scale voltages (V)
Table 3: Signal conditioning gain ranges for 2.5V unipolar or +/- 2.5V @ 2.5VREF bipolar
Table 4: Signal conditioning gain ranges for 1.25V unipolar or +/- 1.25V @ 1.25VREF bipolar
A gain of 50V/V covers most applications and is used in the example later. Should the shunt need to be smaller to reduce power losses or, because of self-heating issues, the gain will be higher, amplifying the error and raising the contribution to the error budget.
The obvious way of measuring voltage drop across a shunt resistor is to use a differential amplifier. Even on the low side this makes sense as it avoids parasitic resistance of the ground plane and ensures that the input can swing to ground. For cost saving reasons, some designers use an operational amplifier plus external resistors in a differential amplifier configuration.
The following section shows what happens on the high and low side analyses. It is assumed that an ideal op amp without offset and drift is used to show how the differential amplifier circuit behaves in both cases. TINA TI is used as the simulation tool to perform worst case scenarios and component tolerances are set to maximum values.
The following circuit can be used to investigate the CMRR in a differential amplifier configuration:
Figure 3: Differential amplifier test circuit for common mode
Figure 4: Output variation vs. common mode ” all R 1%
The CMRR depends on two parameters – the tolerance of the resistors and the gain involved (see Table 5).
Table 5: CMRR for a differential amplifier
Differential amplifiers are very sensitive to common mode voltages as soon as resistor tolerances are taken into account. The gain error of such a circuit is much less then than the influence of the common mode error for high side measurement.
The following example gives an idea of the influence of the resistors using the following values for simulation:
2mOhm shunt with no tolerance
ideal Op Amp
I = 25A (resulting in a shunt full scale voltage = 50mV)
VCM = 0V to 40V
diff amp gain = 50V/V
VREF = 2.5V
Figure 5: Differential amplifier on high side
Figure 6: Output variation – 40VCM – all R 1% – gain 50V/V
Figure 7: Output variation – 40VCM – all R 0.1% – gain 50V/V
Figure 8: Output variation @ I = 0A – all R 1%
As can be seen, the output offset is determined not only by the resistor tolerances, but also by the common mode voltage itself. According to the simulation with 1% resistors, the resulting CMRR (50V/V) = 28.2dB. This means the influence of the resistors only leads to an error of 37% @ 24VCM and 18.5% @ 12VCM. This circuit is unusable. With 0.1% resistors a CMRR (50V/V) = 48.2dB is achieved, leading to an error due to resistor tolerance of only 3.7% @ 24VCM and 1.85% @ 12VCM. Although this approach might be acceptable, we have not yet taken gain error, any drift and op amp errors with offset and offset drift into account. The gain error is 2% for 1% resistors or, 0.2% for 0.1% resistors and a good op amp with less then 500uV offset including drift over temperature will be another 1%. For the error budget using the shunt of 1% with TCR of 0.12% the result is:
Error (24VCM) = Shunt + TCR + Error OP + Gain Error R(0.1%) + CM Error R(0.1%)
Error (24VCM) = 1% + 0.12% + 1% + 0.2% + 3.7% = 6.02%
Error (12VCM) = Shunt + TCR + Error OP + Gain Error R(0.1%) + CM Error R(0.1%)
Error (12VCM) = 1% + 0.12% + 1% + 0.2% + 1.85% = 4.17%
To be able to realise a differential amplifier on the high side, the op amp must be capable of swinging to the high side common mode at its input. With a gain >> 1 there is no longer any advantage to the differential amplifier of dividing the VCM by the input voltage divider, as the circuit's inputs cannot swing wide enough outside the supply rails anymore. This makes the design much more complicated. A wide supply range op amp capable of swinging close to the positive rail, especially at high gains is needed. This normally results in the positive op amp supply voltage rail being higher than the common mode voltage of the shunt.
When considering 0.1% resistors required for midrange accuracy on the high side, the discrete diff amp approach for high side measurement is no longer attractive and this is precisely why current shunt monitors with much better performance at a lower total cost, such as the INA19x and INA210 families from Texas Instruments, have been developed. These components offer high common mode and can swing above and sometimes below their supply rails using no external components other than the shunt resistor.
Common mode voltage has no influence when measuring the low side. In this analysis similar values are used as in the previous example.
Figure 9: Differential amplifier on low side
Figure10: Output variation – all R 1% – gain 50V/V
Figure 11: Output variation – all R 0.1% – gain 50V/V
The resulting offset error comes from the VREF (2.5V) * 2% = +/-50mV or VREF (2.5V) * 0.2% = +/-5mV. The gain error is 2% for 1% resistors or 0.2% for 0.1% resistors. This means at full-scale 5V, the error due to the resistors amounts to 3% or 0.3%. This might be acceptable if care is taken when selecting the operational amplifier. For instance, a 1mV input referred offset, including offset drift of the operational amplifier in relation to the 50mV shunt full-scale voltage, leads to another 2% error. A good choice is an operational amplifier with low offset and offset drift like the OPA333 also from Texas Instruments or, even better with auto zero technology like the OPA335. These permanently cancel offset and offset drift during operation.
Looking at a shunt of 1% tolerance and TCR of 0.12% total error budgets are:
Error (1%) = Shunt + TCR + Error OPA350 + Gain Error R(1%) + Offset Error R(1%))
Error (1%) = 1% + 0.12% + 1.48% + 2% + 1% = 5.6%
Error (0.1%) = Shunt + TCR + Error OPA335 + Gain Error R(0.1%) + Offset Error R(0.1%)
Error (0.1%) = 1% + 0.12% + 0.016% + 0.2% + 0.1% = 1.436%
The approach with 1% resistors might be acceptable if a low accuracy can be tolerated, but the higher accuracy with 0.1% resistors is usually preferred. However, this kind of solution requires a low offset, low drift operational amplifier in combination with accurate and low drift resistors, which might be expensive.
Alternatively, current shunt monitors requiring no external circuitry and specifying the complete signal path, excluding only the shunt resistor, such as TI's INA19x or INA21x families can be used. Both can sense shunt voltage drops at wide common mode independently of the supply voltage, which determines the usable output voltage swing only. INA19x, which employs a different internal approach from that of a differential amplifier, operates at above and below the supply voltage rails. INA21x features auto zero technology and can operate above the positive supply voltage rail, plus includes the negative rail.
For high side and low side current measurement, the shunt measurement method is often used. The shunts tend to have very small resistance to keep power dissipation low. To measure the voltage drop across the shunt resistor, a differential amplifier approach is widely used to enable a wide common mode range. One drawback of the differential amplifier, however, is a strong requirement for resistor matching, which is essential for the common mode rejection and the resulting accuracy. Small shunts require high gain, which amplifies this effect as well as offset and offset drift of the active components, which are the other important requirements for precise current measurement.
Especially when measuring the high side, results are affected by the high common mode. Using external discrete resistors may lead to very high errors unless low tolerance and low drift components are used, but that increases total solution cost dramatically. To overcome these problems, dedicated current shunt monitoring devices have been designed that use highly matched silicon resistors in combination with low offset low offset drift operational amplifiers or, even auto zero amplifiers.
Author: Lutz Nauman, Texas Instruments