What is low-side current sensing?

In Signal Chain Basics #93: How to Maximize Low-Side Sensing Performance, I discussed various options for low-side current sensing. A quick reminder, low-side current-sensing is where the sense element, or shunt resistor, is placed in series with the load between the load and ground (**Figure 1** ). This is the most common method to measure current.

**Figure 1**

The drawbacks to low-side sensing are disturbances to the system load’s ground potential and the inability to detect load shorts to ground. Low-side sensing is desirable because the common-mode voltage is near ground. As discussed in article **#93** , while there are many alternatives for low-side current measurements (**Table 1** ), current-shunt monitors (or current-sense amplifiers) offer significant accuracy advantages, lower power consumption, smaller footprints, enable the use of smaller shunt resistors, and in many cases at equal or lower cost.

**Table 1**

**Current-sense amplifier accuracy advantages**

To understand how a current-sense amplifier maximizes the achievable accuracy, let us look at the total error in current measurement per the root-sum-square error calculation:

I will focus on the two primary error contributors [1]: VOS and gain%.

The error due to offset voltage (*e _{VOS-MAX } * ) is simply the signal-to-noise ratio (SNR) of the voltage generated by the minimum current flow through the shunt resistor to the worst case offset voltage found in the device data sheet. Why the minimum current, let us look at how (

*e*) is calculated (Equation 2):

_{VOS-MAX }Now compare a common op amp, LMV341, and a moderate accuracy, low-cost current-sense amplifier INA199, and their respective errors due to V_{OS} . The op amp has a VOS maximum value at 25^{o} C of 4.0 mV, while the current-sense amplifier is at 150 μ V. If we assume a minimum current 10A and a 4 mΩ shunt resistor, the op amp results in a 10 percent error, while the current-sensing amplifier is only 0.4% – a clear improvement.

The second most critical error source is due to gain error, which is just a straight forward percentage adder. For an op amp implementation (**Figure 2** ), the gain is set by two external resistors (R_{F} and R_{G} ). The resistor tolerance sets the gain error. The designer then has to make a tradeoff on cost of the external resistors versus the desired accuracy – more accurate resistors cost more! For a current-sense amplifier, the gain error is expressed as a simple percentage in the data sheet.

**Figure 2**

What if the error level in our first example is acceptable, and the cost to implement the op amp is less than the current-sense amplifier? Since the current-sense amplifier also saves system power by using a lower value shunt, it may be the better choice. Using the same operating guidelines above, we can reduce the shunt value to 0.15 mΩ. Since we are making the error calculations at the minimum current, let us look at the power savings at the maximum current of 50A. Using the shunt value for both implementations that can achieve 10 percent error at the minimum current, the op amp implementation adds 10W to the system’s power consumption, versus only adding 0.375W to the system power budget with the current-sense amplifier.

**Implementation costs**

Let us look at the rough cost of implementation. To simplify the comparison, assume that the supply voltage, necessary bypass capacitors, and shunt resistor cost are the same for each. The only items left to consider are the signal path to amplify the signal.

Based on **Figure 2** , an op amp implementation’s BOM consists of the device and two resistors. Now assume those resistors are standard five percent and cost rough $0.01, combined. For our example, if we add the listed price for the op amp from the TI website ($0.20), implementation cost is ∼$0.21. To increase gain accuracy, if we changed the resistors to 0.5%, the total cost could jump as high as $0.24.

The current-sense amplifier implementation shown in **Figure 1** requires no external components, so the TI web price of $0.35 is the total cost of implementation, and offers better than 0.1% gain accuracy.

Finally, consider the effect of temperature on these measurements and costs to implement. Many current sense amplifiers feature a zero-drift architecture that reduces the change in V_{OS} performance over a wide temperature range. Additionally, the integrated gain resistors are precisely matched and offer very low temperature coefficients (TEMPO). The INA199s TEMPCO is 10 PPM/^{o} C. This combination minimizes the temperature effect on accuracy for a current-sense amplifier implementation. Many op amps feature a similar zero-drift architecture that helps with the VOS error over temperature. However, the external gain resistors cannot be precisely matched and typically have fairly large TEMPCOs (>200 PPM/^{o} C). The cost to buy low-TEMPCO resistors drastically increases to the point that the cost of an op amp implementation will be equal to or greater than that of a current-sense amplifier.

**Summary**

When you compare a current-sense amplifier to an op amp for low-side current sensing, the current-sense amplifier offers:

- Significantly improved accuracy: >96% (
*e*reduction in the example_{VOS } - Reduced power consumption for the same error levels: >96% in the example
- Reduction in gain error and temperature drift of the gain network

While this improvement can potentially increase the BOM, depending on temperature drift requirements, the relative increase in cost is less than half the improvement in performance.

Join us next time when our focus will be on high-speed amplifiers.

**References**

- Jason Bridgmon, Getting Started with Current Sense Amplifiers, Session 5: Understanding Different Types of Error in Current Shunt Monitor Designs, TI Training
- Download the INA199 data sheet

Be careful. There is no “ground”. PCB resistance can easily add to the sense resistor as I show in my flat impedance webinar. Even some of the evaluation boards we tested showed 30% error due to PCB effects using a 10 mohm sense resistor. The smaller the resistor the larger the error.

Right, both + and – inputs to the sense amplifier must have direct Kelvin connections to the sense resistor. And if a pcb layout is re-worked, be sure to check that the layout person did not move a sense resistor and remove the Kelvin connection. That happened to me once, now I always check for this.

Many years ago a colleague was trying to make a product intrinsically safe using the same topology although IIRC the resistance was a couple of ohms, not milliohms. It took quite a bit of work to stop the circuitry oscillating.

I've used low-side sensing for only one system, and Kelvin connections were employed. However, the failure mode was that the negative side of the shunt resistor could be disconnected from ground, and the full Vbus=120VDC would be applied to the opamp or sense amplifier. We used fully isolated (floating) amplifiers to prevent catastrophic system failure.