Baker's Best Blog

Improve EMI rejection with a simple front-end amplifier

Electromagnetic interference (EMI) is a part of our lives, engineer or not. The proliferation of electronic solutions is a good thing as electronic devices bring comfort, safety, and health to our lives. However, all these good things continue to clutter our transmission air space. The best defense to this interference is to nip this problem in the bud with solutions specifically designed to run obstruction to the interference. This blog shows how to quantify and quickly solve EMI issues in your sensor circuits.

EMI signals come from various sources. These sources include the standard electronic devices around us; cars, trucks, and heavy equipment are inherent generators of EMI signals. And something near and dear to our 21st-century hearts is the cell phone. The cellphone offers the convenience of communicating with friends, family, and business associates while walking down the street. Nevertheless, this gadget has the potential to produce EMI signals that interfere with our electronics. So, leave the cellphone outside your lab.

Load cells and EMI errors

A great candidate for EMI signals disruption is with sensors that have extremely low voltage or current output signals. The rectification of an EMI signal can appear as intermittent DC voltage and current offsets. For instance, load cell bridges produce a class of analog signals that present complex acquisition problems. The load cell signal output can be in the sub-millivolt region, and if you are interested in precision, the amplifying electronics complicate measurement activities.

Take the example of four-element load cell sensor model configured in a resistive bridge, as shown in Figure 1. Assuming a load is applied to the sensor, the voltage at the sensor’s output produces small signals, tens of mV maximum, appearing between the mid-points of the two resistive legs.

Figure 1 The output of a load cell bridge drives a two op-amp instrumentation amplifier. Source: Bonnie Baker

This design configuration is typically called the two op-amp instrumentation amplifier. In this discrete design, TI’s OPA2187 dual amplifiers have good bandwidth and over temperature matching. TI’s REF2025 2.5-V reference level shifts A2’s output to a central supply voltage. This instrumentation amplifier reduces source impedance mismatch problems by using the high impedance of the operational amplifier’s non-inverting input.

The transfer function of this circuit is equal to:

VOUT = (VIN+ – VIN-)(R4 (1 + ½(R2/R1 + R3/R4) + (R2 + R3)/RG)/R3) + VCM (R4(R3/R4 – R2/R1)/R3) + 2.5V

When R1 = R4 and R2 = R3, the transfer function becomes:

VOUT = (VIN+ – VIN-)(1 + R1/R2 + 2R1/RG) + 2.5V

With matched resistors, the circuit gain changes with one resistor, RG.

For example, a full-scale 5 V in 16-bit analog-to-digital converter (ADC) yields 1 LSB of 5V/216 or 76.29 µV. An amplifier with an offset voltage of 500 µV is well above 1 LSB. To maintain linearity and avoid quantization errors, select a precision amplifier that yields ½ LSB. A zero-drift amplifier such as the OPA2187 has 10 µV of offset voltage and 0.001 µV/℃ of offset voltage drift.

EMI-hardened op amp advantages

Zero-drift amplifiers—such as OPA2187 dual 0.001 zero-drift op amps—have low voltage offsets and low 1/f noise. The preventative measures of EMI problem avoidance include filtering, shielding, and proper grounding. The OPA2187 has EMI and radio frequency interference (RFI) filtered inputs. A simple low-pass passive filters RC filter, whether at the input or output, and can affect the dynamic performance of the amplifier. The most effective way to reject EMI and RF signals is to use the integrated approach.

The closely matched silicon integrated filters in OPA2187 reduce signal path errors that feed the ADC. The EMI rejection ration (EMIRR) plots in product datasheet provide a better understanding of how EMI-hardened amplifiers reduce errors.

For example, suppose a non-EMI-hardened op amp that provides 50 dB of EMI rejection has a gain of 100. This amplifier interfaces with a 5-V full-scale 16‑bit ADC.

At the amplifier’s input, there is a –20 dBV or 0.1 V RF signal. A computation yields a 0.32 mV EMI error at the input or 0.1 V / 10(50 / 20). The EMI error of 0.32 mV times a gain of 101 produces a 31.9 mV error. With a 5‑V full-scale voltage range and a 16-bit ADC, the LSB size is 76.29 µV. Approximately, a 419 digital count loss equals the division of the 31.9 mV EMI error by 76.29 µV. A zero-drift amplifier, such as the OPA2187, provides ~100 dB of EMIRR at 1 GHz (Figure 2).

Figure 2 The plot shows EMIRR ratio versus frequency for the OPA2187 amplifier. Source: Texas Instruments

A new computation error yields a 1 µV EMI error at the input or 0.1 V / 10(100 / 20). The EMI error of 1 µV times a gain of 101 produces a 101 µV error. Approximately, a 1.4 digital count loss equals the division of the 101 µV EMI error by 76.29 µV.

The above design examples show how EMI can creep into your circuits without a warning. In sensor circuits, the EMI manifests itself as voltage offset and offset current. The blog has provided a computation strategy to assess the damages upfront. Pull out your pencil and paper to quantify the potential EMI sources.

The best defense against this interference is always filtering, shielding, and proper grounding techniques to nip this problem in the bud. Then apply specifically-designed devices to run obstruction to the interference. The blog has also shown how to quantify and quickly solve EMI issues in your sensor circuits.

Bonnie Baker is a seasoned analog, mixed-signal, and signal chain professional and electronics engineer. Baker has published and authored hundreds of technical articles and blogs in industry publications. She is also the author of the book “A Baker’s Dozen: Real Analog Solutions for Digital Designers” as well as coauthor of several other books.

Related Content

0 comments on “Improve EMI rejection with a simple front-end amplifier

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.