SIGNAL CHAIN BASICS #53: Properly scaled filter components improve noise attenuation

(Editor's note : Signal Chain Basics is an ongoing (and popular) series; you can click here for a complete, linked list of all previous installments of the series.)

In DC to low-frequency sensor signal-conditioning applications, relying on the common-mode rejection ratio (CMRR) of an instrumentation amplifier to provide robust noise rejection in harsh industrial environments is rarely sufficient.

To avoid the propagation of unwanted noise signals, it is critical to properly match and scale the components in a low-pass filter at the inputs of the instrumentation amplifier. Ultimately, this enables the internal electromagnetic interference/radio frequency interference (EMI/RFI) filtering and CMRR to provide the remaining noise attenuation needed to achieve an acceptable signal-to-noise ratio (SNR).

To illustrate, consider the low-pass filter implementation shown in Figure 1 . The resistive sensor is connected differentially to a high-impedance instrumentation amplifier through a low-pass filter network comprised of RSX , and CCM . Ideally, if CCM in each input leg is perfectly matched, the amount of noise that is common to both inputs will be attenuated symmetrically before reaching the inputs of the INA.

Figure 1: Common-mode input filtering.

 With a perfect match in common-mode filter capacitor (Ccm ), noise is almost completely attenuated. This is shown in the TINA SPICE simulation in Figure 2 by injecting a 100-mVpp, 100-kHz common-mode error signal into the INA333 inputs.


Figure 2: Example simulation of perfect input-RC match for common-mode filtering using INA333.

 The problem with this approach is that off-the-shelf capacitors have a typical tolerance of five to 10 percent, which means that if CCM in each leg is mismatched in opposite directions, the overall differential tolerance can be as much as 20 percent. This capacitive mismatch is better represented by Figure 3 , which also illustrates the entry of common-mode noise (eN ) at the output of the resistive sensor.


Figure 3: Common-mode filtering with RC mismatch and common-mode noise injection.

 This mismatch at the inputs (?C) creates an error in cutoff frequency that causes common-mode noise, eN , to pass differentially into the inputs of the INA, which is then gained up to the output as an error voltage. The amount of common-mode noise that gets passed to the inputs is shown in Equations 1 through Equation 3 :


Assuming that the sensor signal Vsensor is much lower in frequency than the noise cutoff of each common-mode filter (i.e., fC ≥ 100•fsensor ) and RS1 = RS2 , then the amount of common-mode noise signal (eN ) that gets converted to a differential noise signal (eIN ) and becomes part of VIN is given in Equation 4 :


Figure 4 further illustrates the errors that can propagate to the output by injecting a 100-mVpp, 100-kHz common-mode error signal into the INA333 with a 10 percent RC mismatch on the filter cutoff frequency of 1.6 kHz:


Figure 4: Simulation of output error of INA333 (in gain of 101) due to common-mode filter RC mismatch.

 A better and more common approach to input filtering is shown in Figure 5 with the small improvement of adding a differential capacitor, Cdiff , across the inputs of the instrumentation amplifier.

Figure 5: Adding differential capacitor (Cdiff ) for improved common-mode noise attenuation.

Adding this capacitor does not solve this problem completely as Cdiff must be scaled with the following two criteria in mind:

  1. The differential cutoff frequency must be large enough to be sufficiently out of the way of the signal bandwidth and allow sufficient filtering settling. 
  2. The differential cutoff frequency must be small enough to attenuate common-mode noise to an acceptable level such that the instrumentation amplifier’s CMRR can provide the remaining attenuation needed to achieve an acceptable SNR. A good rule of thumb is given in Equation 5 :

Figure 6 shows a plot of VinP and VinN versus frequency with no Cdiff and with Cdiff = F. Notice that with no differential capacitor there is a difference in the INA333’s output magnitude. This difference gets amplified to the output as noise that ultimately degrades the SNR. With Cdiff = F, the difference between VinP and VinN is minimal.

Figure 6: Plot of VinP and VinN with Cdiff = 0 and Cdiff = 1 F.

Figure 7 shows this improvement in overall noise performance at the output of the INA333 by adding with Cdiff = F.

Figure 7: Simulation of improved noise filtering using Cdiff in INA333.

In summary, a low-pass filter placed in front of an instrumentation amplifier should have a differential capacitor scaled at least 10 times higher than the common-mode capacitor. This significantly will improve the effectiveness of the filter by reducing the effects of mismatching in Ccm , which can cause common-mode noise to become differential noise.

Join us next time when we will address the intricacies of I2 S clocks in a master/slave system.


  • Download a datasheet for the INA333:
  • Download free software from TI:
  • Learn more about SPICE:

About the author
Matthew Hann is an applications manager in the precision analog group at Texas Instruments.

0 comments on “SIGNAL CHAIN BASICS #53: Properly scaled filter components improve noise attenuation

Leave a Reply

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