The fully isolated circuit shown in Figure 1 monitors individual channels of −48 V current with better than 1% accuracy. The load current flows through a shunt resistor located outside the circuit. The shunt resistor value should be chosen so that the shunt voltage is approximately 50 mV at maximum load current.

The AD7171 measurement results are provided as digital codes through a simple 2-wire SPI-compatible isolated serial interface. Isolation is provided by the ADuM5402 quad-channel isolator . In addition to isolating the output data, the ADuM5402 digital isolator provides an isolated +3.3 V supply to the circuit.

This combination of devices enables an accurate high-voltage negative rail current sensing solution with low component count, low cost, and low power consumption. The measurement accuracy mainly depends on the resistor tolerance and the accuracy of the bandgap reference voltage source, with typical values ​​being better than 1%.

This circuit is designed for a full-scale shunt voltage of 50 mV at maximum load current IMAX. Therefore, the shunt resistor value is R _SHUNT = (50 mV)/(I _MAX ).

The "ground" of the op amp stage is connected to the common-mode source voltage (−48 V). The voltage for the op amp stage is provided by a "floating" 5.6 V Zener diode that is biased to approximately 2 mA, eliminating the need for a separate power supply. Without modifications, this circuit has a source voltage range of −60 V to −10 V.

U1A amplifies the shunt voltage 49.7 times, where G = 1 + R3/R2. The offset voltage of the zero-drift amplifier ADA4051-2 is very low (15 μV maximum) and does not contribute significantly to the measurement error. A full-scale shunt voltage of 50 mV produces a full-scale output voltage of 2.485 V from U1A (referenced to the common-mode source voltage).

U1B's feedback loop contains an N-channel MOSFET transistor with a large VDS breakdown voltage (70 V), which applies U1A's output voltage across resistor R5, resulting in current flowing through R6 and R7. The 2.485 V full-scale voltage from U1A produces a full-scale current of 0.498 mA, which produces a 2.485 V full-scale voltage across resistor R7. The voltage across R7 is applied to AIN− of the ADC. Resistor R6 and Schottky diode D2 provide input protection for the AD7171 when the MOSFET is shorted.

Note that the supply voltage for the ADR381 , AD7171, and floating Zener diodes is provided by the isolated supply output (+3.3VISO) of the ADuM5402 quad isolator.

The reference voltage of the AD7171 is provided by the precision bandgap reference voltage source ADR381. The ADR381 has an initial accuracy of ±0.24% and a typical temperature coefficient of 5 ppm/°C.

Although both the AD7171 VDD and REFIN(+) can operate from an isolated 3.3 V supply, using separate voltage references provides greater accuracy. The lower isolation voltage limit is 3 V, so a 2.5 V reference voltage must be used to provide adequate headroom.

The input voltage to the AD7171 ADC is converted to an offset binary code at the output of the ADC. The ADuM5402 provides isolation for the DOUT data output, SCLK input, and PDRST input.

After the isolation circuit, the code is processed in the PC using the SDP hardware board and LabVIEW software.

The graph in Figure 2 shows that the circuit under test achieved an error of 0.3% over the entire input voltage range (0 mV to 50 mV). In addition, the ADC output code recorded by LabVIEW was compared with the ideal code calculated based on the ideal system.

In order to calculate this ideal code, several assumptions must be made regarding system performance. First, the op amp stage must amplify the input signal exactly 49.7 times. This value varies by 2% in the worst case, depending on the resistor tolerance (1%). Second, assume that the current sink resistor (R5) is exactly the same as the ADC input resistor (R7). In this circuit, the tolerance of these resistors is 1%. Since they have the same value, the matching accuracy is probably better than 1%. It is also possible to use lower tolerance resistors, which will increase the accuracy and cost of the circuit.

There are several other components mounted on the PCB that are not critical to the functionality or performance of the circuit, but are necessary to ensure the safety of the user and the hardware. For example, if Q1 breaks down or shorts, the ADC, SDP board, and user PC may be damaged by a large negative voltage. Safety components include passive components R6 and D2 to protect the AD7171, and a quad digital isolator ADuM5402 to protect the circuitry on the SDP board and the user PC.

PCB layout considerations

In any circuit where precision is important, power and ground return layout on the circuit board must be carefully considered. The PCB should isolate the digital and analog parts as much as possible. This PCB is made of four-layer boards stacked with large area polygons for the ground layer and power layer. For a detailed discussion of layout and grounding, refer to Tutorial MT-031 ; for information on decoupling techniques, refer to Tutorial MT-101.

The supplies for the AD7171 and ADuM5402 should be decoupled with 10 μF and 0.1 μF capacitors for proper noise rejection and ripple reduction. These capacitors should be as close as possible to the corresponding device, and the 0.1 μF capacitors should have low ESR values. For all high frequency decoupling, ceramic capacitors are recommended.

The isolation gap between the primary and secondary sides of the ADuM5402 should be carefully considered. The EVAL-CN0188-SDPZ board maximizes this distance by pulling back the polygon or device on the top layer and aligning it with the pins on the ADuM5402.

Power traces should be as wide as possible to provide a low impedance path and reduce the effects of glitches on the power lines. Clocks and other fast-switching digital signals should be digitally shielded from other devices on the board.

For a complete design support package for this circuit note, including board layout, see http://www.analog.com/CN0188-DesignSupport .