Industrial

# The accuracy and cost benefits of precision

14th August 2019
Joe Bush

As Farhana Sarder of ON Semiconductor explains, general purpose op amps aren’t all purpose op amps.

We often find customers using general purpose op amps like the LM321 for current sensing applications. This is one of the legacy op amps that have been around for decades. These legacy op amps are low cost, and they’re used in countless applications. However, sometimes the same customers come back to ON Semiconductor saying that these op amps are failing in their current sensing circuits. When we take a look at the returned op amp units, they’re working just as expected. So what went wrong?

Just because an op amp is ‘general purpose’ doesn’t mean its ‘all purpose’. Current sensing applications need precision. Current sensing is typically used in applications for power management and overcurrent protection. Imagine a world where we didn’t have precision. Your phone’s battery gauge might read eight percent when your battery is really about to die. You might design your overcurrent circuit to trigger at 100A, but then you find that the protection circuit doesn’t kick in until 150A, and all of your downstream devices are damaged. This is the difference between general purpose and precision.

A precision op amp is all about input offset voltage. It also will have better specs for CMRR and PSRR, but both of these parameters can be realised as input offset voltage that changes with common mode voltage or power supply voltage. So, what is input offset voltage? It’s an inherent offset at the input of each and every op amp, and it’s due to slight mismatches in the input transistor pair that arise from the manufacturing process. In school we learn that the ideal op amp has zero input offset voltage, but we know that’s not the case living in the real world.

## The reality

A legacy general purpose op amp like the LM321 has VOS =±7mV max, and a modern general purpose op amp like the NCS20071 has VOS = ±3.5mV max. This max specification is based on a distribution centred near zero. That means that most of the time a randomly chosen part will exhibit near-zero offset.

You could be convinced that your prototype circuit works perfectly with a common LM321, but then when the circuit goes into mass production, you may find that you have a significant percentage of failures. This is because there is part-to-part variation due to the manufacturing process and some parts will be near the limits. You should always design your circuit for the maximum input offset voltage. Occasionally, ON Semiconductor see customers forgetting to check the circuit for the worst case limits: input offset voltage limits, CMRR limits, resistor network tolerances, temperature effects, etc.

Figure 1. Comparison of input offset voltages and resulting output offset error. The amplifiers with 7mV and 3.5mV input offsets have noticeable output offset error

Compare the LM321 and NCS20071 general purpose op amps to the new NCS21911 precision op amp, which has a maximum offset of VOS = ±25µV (microvolts) thanks to its chopper-stabilised architecture. How much difference does offset voltage really make? Let’s consider a situation with a fixed 50mV shunt drop, as shown in Figure 1. We can look more closely at the example with VOS = 7mV in Figure 2.

By opting for a precision op amp like the NCS21911, the error contribution from input offset voltage will become almost negligible in this circuit example. Not only does the output accuracy improve, but there is even some margin to decrease the size of the sense resistor and still maintain the required accuracy.

Since the low offset voltage allows you to decrease the sense resistor value while maintaining the same accuracy, as seen in Figure 3, efficiency can be greatly improved. What happens when the sense resistor size is decreased? There is less power consumed through the sense resistor which means that a lower wattage and lower cost resistor can be used, and the physically smaller sense resistor ends up taking up less space on the PCB. The overall efficiency of the system improves and less power is wasted.

## Application differences

In many applications, the load current through the sense resistor is variable. Sometimes when customers try to take a current measurement near 0A, they find that the error has increased significantly; this is normal and should be expected. As the current goes down to zero, the error percentage goes to infinity. This current sensing circuit is designed to measure current; it’s not designed for accurate measurement when there is no current. Figure 4 shows how the accuracy improves as the current increases. Notice how the error changes due to input offset voltage. The 25µV offset of NCS21911 allows relatively accurate measurements even when the sense voltage decreases.

What seems like small improvements in efficiency and accuracy can add up to savings on the bill of materials (BOM), the cost of the PCB, and the electricity bill. While opting for the less expensive op amp might save some money up front, consider that the system level savings might ultimately be to your advantage with a reasonably priced precision op amp.

Figure 2. Low-side current sensing and the input offset voltage contribution to output error

There are many applications where a general purpose op amp will work just fine. Even a legacy LM321 can work in a current sensing application where the circuit has been designed accordingly. Keep in mind that you should expect a relatively higher output error. Alternatively, the sense resistor should be sized larger to get a voltage drop that’s sufficiently larger than the input offset voltage.

For low side current sensing, switching to a precision op amp improves the accuracy and the system efficiency. The NCS21911 precision op amp has a standard pin-out, which makes it a simple drop-in replacement for general purpose op amps like LM321 and NCS20071.

Figure 3. Comparison of input offset voltages and resulting shunt drop with fixed accuracy requirement. The smaller shunt drop results in improved efficiency

Figure 4. Error due to input offset voltage