3

I am using an AD8217 high side current sensor to measure currents up to 20A. It works as expected when testing with lower currents and sense resistors of 1\$ \Omega \$ and 0.3\$ \Omega \$ and currents of 660mA and 130mA.

For the eventual 20A, I will need to use a 5m\$ \Omega \$ sense resistor. The output of the AD8217 is not what I expect using this resistor and the 660mA or 130mA currents.

At 130mA, I expect an output of 13mV but I measure 34mV. At 660mA, I expect 66mV but measure 115mV.

Directly measuring the differential voltage across the sense resistor looks right (e.g., 3.3mV with 660mA which should produce an output of 20 x 3.3mV = 66mV).

The application note (https://www.analog.com/en/analog-dialogue/articles/high-side-current-sensing-wide-dynamic-range.html) says that a differential voltage across the sense resistor of at least 1mV is fine. I don't think that the AD8217 offset voltage or bias current can account for what I am seeing.

Thanks for any ideas.

This is the circuit I am using with a 12.5V battery, either 18\$ \Omega \$ or 100\$ \Omega \$ load resistors and 0.3\$ \Omega \$, 1\$ \Omega \$ or 0.005\$ \Omega \$ sense resistors.

enter image description here

Mitch
  • 133
  • 4
  • 1
    What voltage are you seeing between the GND pin of the AD8217 and the GND measurement point you have used? Is the GND on the 8217 linked to the GND point on the power circuit? – Andy aka Apr 23 '19 at 14:00
  • 1
    What kind of load do you have at the output of the AD8217? It has a limited ability to sink current close to ground (~40 mV @ 1 mA). – Dave Tweed Apr 23 '19 at 14:18
  • @Andy aka: I don't think that should matter if I am measuring OUT relative to the AD8217 ground. – Mitch Apr 23 '19 at 15:11
  • @Dave Tweed: The load is only the meter. But what you're suggesting may be a factor since the minimum voltage out is 10mV. – Mitch Apr 23 '19 at 15:16
  • Insufficient data. Try a triangular wave to a transistor active load and view output linearity, gain, offset error – Tony Stewart EE75 Apr 23 '19 at 16:17

1 Answers1

3

Did you use a Kelvin connection to connect the 5m\$ \Omega \$ to the AD8217?
If not, the measurement is not only the voltage drop across the resistor itself, but also across the contacts/traces as well.

enter image description here

In the upper picture, the connection to the AD8217 has been made with the thin traces. However, the thick trace will have an impedance (resistance and inductance). Also the solder joints (or whatever connection you use) to the shunt resistor will have contact resistance.
In case of using 1\$ \Omega \$ or 0.3\$ \Omega \$ these contact resistance could probably be neglected or be incorrectly dismissed as being AD8217 offset voltage or bias current.
When using a 5m\$ \Omega \$ shunt resistor, you can certainly not neglect the above mentioned impedances.

The impedances cause additional voltage drops (I represented them as \$V_{TC}\$), which is current dependent, which seems to match your findings.

Using a Kelvin connection as shown in the lower picture will prevent measuring these voltage drops.

Huisman
  • 10,694
  • 2
  • 20
  • 40
  • 1
    Using a Kelvin connection with measurement wires connected close to the body of the resistor improved the situation considerably. Now I measure 15mV vs. 13mV expected and 72mV vs. 66mV. I suspect that these differences can be reduced with more careful wiring. Thanks. – Mitch Apr 23 '19 at 15:08