Catherine Chang, Linear Technology
Design Note 1045
Precision high side measurement of microamp currents requires a small value sense resistor and a low offset voltage, ultralow-power amplifier. The LTC2063 zero-drift amplifier has a maximum input offset voltage of just 5 μV and draws just 1.4 μA making it a great choice for building a complete precision high side current sense circuit, as shown in Figure 1.
|Figure 1.||Precision High Side Current Sense Circuit Based on the LTC2063 Zero-Drift Amplifier.|
This circuit uses only 2.3 μA to 280 μA of supply current to sense currents over a wide 100 μA to 250 mA dynamic range. The exceptionally low offset of the LTC2063 allows this circuit to work with only 100 mΩ of shunt resistance, limiting the maximum shunt voltage to only 25 mV. This minimizes power loss on the shunt resistor and maximizes power available to the load. The LTC2063’s rail-to-rail input allows this circuit to operate with very small load current where input common mode is almost at the rail. The integrated EMI filter of the LTC2063 protects it from RF interference in noisy conditions.
The voltage output of this circuit for a given sense current is:
A critical specification for a current sense solution is the zero point, or equivalent error current at the input for the output produced when no sense current is present. The zero point is generally determined by the input offset voltage of the amplifier divided by RSENSE. The LTC2063’s low input offset voltage, of typical 1 μV, maximum 5 μV, and low typical input bias and offset currents of around 1 pA, allow for a zero point input-referred error current of only 10 μA (1 μV/0.1 Ω) typically, or 50 μA (5 μV/0.1 Ω) maximum. This low error allows the sense circuit to maintain its linearity down to the lowest current in its specified range (100 μA), without plateauing due to loss of resolution, as seen in Figure 2. The resulting input current to output voltage plot is linear over the entire current sense range.
|Figure 2.||No Plateau at the Low End, Down to 100 μA ISENSE.|
Another source of zero point error is the output PMOS’s zero gate voltage drain current, or IDSS, a parasitic current that is present for nonzero VDS when the PMOS is nominally turned off (|VGS| = 0). A MOSFET with high IDSS leakage will produce a non-zero positive with no ISENSE.
The transistor used in this design, Infineon’s BSP322P, has an upper-bound IDSS of 1 μA at |VDS| = 100 V. As a good estimate for the typical IDSS of the BSP322P in this application, at room temperature, with VDS = –7.6 V, IDSS is only 0.2 nA, resulting in just 1 μV error output, or equivalent 100 nA input current error, when measuring 0 A input current.
The LT1389-4.096 reference, along with the bootstrap circuit composed of M2, R2, and D1, establishes a very low power isolated 3 V rail (4.096 V + VTH of M2, typically –1 V) that protects the LTC2063 from seeing its absolute maximum supply voltage of 5.5 V. Although a series resistance could suffice for establishing bias current, using transistor M2 allows for much higher overall supply voltages while also limiting current consumption to a mere 280 μA at the high end of the supply range.
The LTC2063’s input offset voltage contributes a fixed input-referred current error of 10 μA typical. Out of 250 mA full-scale input, the offset results in only 0.004% error. At the low end, 10 μA out of 100 μA is 10% error. Since the offset is constant, it can be calibrated out. Figure 3 shows that total offset from LTC2063, unmatched parasitic thermocouples, and any parasitic series input resistances is only 2 μV.
|Figure 3.||VIN to VOUT Conversion on Minimum Supply 4.5 V
for Entire ISENSE Range. Output Offset of 200.7 μV,
when Divided by 100.05 V/V Voltage Gain, Implies
RTI Input Offset of 2 μV.
The gain shown in Figure 3, 100.05 V/V is 1.28 greater than the expected gain given by the real values of RLOAD and RIN, or 98.77 V/V. This error may be due to the different temperature coefficients of RLOAD and RIN. The main source of uncertainty in the output of this circuit is noise, so filtering with large parallel capacitors is crucial to reduce noise bandwidth and thus the total integrated noise. With a 1.5 Hz output filter, the LTC2063 adds about 2 μVP-P low frequency, input-referred noise. Averaging the output over the longest possible duration further reduces error due to noise.
Other sources of error in this current sense circuit are parasitic board resistance in series with the RSENSE at the LTC2063 input, tolerance in resistance values of the gain-setting resistors RIN and RLOAD, mismatched temperature coefficients in the gain-setting resistors, and error voltage at the op amp inputs due to parasitic thermocouples. The first three sources of error can be minimized by using Kelvin sense 4-lead sense resistors for RSENSE, and using 0.1% resistance with similar or low temperature coefficients for the critical gain-set path of RIN and RLOAD. To cancel out the parasitic thermocouples at the op amp inputs, R1 should have the same metal terminals as RIN. Asymmetric thermal gradients should also be avoided as much as possible at the inputs.
|Figure 4.||Percent Error Remains Below 1.4% for Entire
Range of Readings.
The overall contribution of all the error sources discussed in this section is at most 1.4% when referenced against full-scale 2.5 V output, as shown in Figure 4.
The minimum supply current required by the LT1389 and LTC2063 is 2.3 μA at the minimum VSUPPLY and ISENSE (4.5 V and 100 μA), up to 280 μA at maximum VSUPPLY and ISENSE (90 V and 250 mA), as shown in Figure 5. In addition to the current consumed by the active components, an output current through M1 also supplied by VSUPPLY is required, proportional to the output voltage, ranging from 200 nA for a 0.1 mV output (for 100 μA ISENSE) to 500 μA for a 2.5 V output (for 250 mA ISENSE). Thus, the total supply current in addition to ISENSE ranges from 2.5 μA to 780 μA. RLOAD is set at 5 kΩ for a reasonable ADC drive value.
|Figure 5.||Supply Current Increases with Supply Voltage
but Never Surpasses 280 μA.
In this architecture, the maximum supply is set by the maximum |VDS| that the PMOS output can withstand. The BSP322P is rated for 100 V, so 90 V is an appropriate operating limit.
This design can drive a 5 kΩ load, which makes it a suitable stage for driving many ADCs. The output voltage range is 0 V to 2.5 V. Since the LTC2063 has rail-to-rail output, the maximum gate drive is only limited by the LTC2063’s headroom. It is 3 V typically in this design, set by the LT1389’s 4.096 V plus the –1 V typical VTH of M2.
Since the output of this circuit is a current, not a voltage, ground or lead offset does not affect accuracy. Thus, long leads can be used between the output PMOS M1 and RLOAD, allowing RSENSE to be located near the current being sensed while RLOAD is near an ADC and other subsequent signal chain stages. The drawback of long leads is increased EMI susceptibility. 100 nF C3 across RLOAD shunts away harmful EMI before it reaches the next stage’s input.
Since the LTC2063’s gain-bandwidth product is 20 kHz, it is recommended to use this circuit to measure signals 20 Hz or slower. The 22 μF C2 in parallel with the load filters the output noise to 1.5 Hz for improved accuracy and protects the subsequent stage from sudden current surges. The trade-off of this filtering is longer settling time, especially at the lowest end of the input current range.
The LTC2063’s ultralow input offset voltage, low IOFFSET and IBIAS, and rail-to-rail input, provide precise current measurements over the entire range of 100 μA to 250 mA. Its 2 μA maximum supply current enables the circuit to run on far less than 280 μA supply current for most of its operating range. Along with LTC2063’s low supply voltage requirements, the low supply current allows it to be powered from a voltage reference with headroom to spare.
You may have to register before you can post comments and get full access to forum.