Part 1: click here
Part 2: click here
Part 3: click here
Part 4: below
Op amp input filtering and EMI-hardened op amps
Part 4 of this series recognizes that, in spite of all efforts to keep the EMI and RFI from corrupting the sensor signals on their way from signal source to the signal conditioning instrumentation. Two methods of mitigating this possibility exist: one is to use common-mode and differential-mode filtering on signal lines, as close to the amplifier as possible, and the other is to substitute EMI hardened amplifiers for ordinary amplifiers. This latter is a relatively recent development. Both will be discussed here.
Filtering at the circuit level: signal line filtering
With or without a cable shield, some level of filtering will generally be required. The extreme sensitivity of analog front-ends, to even a few tens of millivolts of conducted RFI on their input pins, requires plenty of filtering to suppress both common-mode and differential-mode noise components. CM rejection is more bothersome than DM because at higher frequencies, the CMRR of the amplifier degrades due to imperfections in input stage symmetry and a decline in the impedance of the input stage tail-current current source.
Also, the simultaneous presence of CM and DM out-of-band RF noise at the inputs of op amps causes rectification, demodulation and inter-modulation effects owing to the inherent non-linearity of the input stage. The results are large shifts in the DC output voltage and distortion. Once these affects take place, there is no hope of filtering them from the desired signals.
The design of the filter circuit itself--to be capable of providing the required degree of attenuation (or insertion loss) for the unwanted signals--must take into account the source and load impedances as well as the spectra of the wanted and unwanted signals in the conductor. The effectiveness of the filter configuration depends on the impedances seen at either end of the filter network. A simple inductor circuit will give good results--better than 40 dB attenuation--when facing a low impedance circuit, but will be quite useless when looking directly into high impedances, such as the inputs to an op-amp or in-amp. A simple capacitor will give good results at high impedances but will be useless at low ones. Multi-component filters, either RC or LC, will give better results provided that they are configured correctly; the capacitor should face a high impedance and the inductor or resistor a low one.
The distinction between the modes of interference coupling (DM or CM) is crucial for filter design. Filters must be configured appropriately to attenuate these modes.
It is almost never the case that the amplifier inputs present balanced impedances. Because of slight imperfections, the two paths (even when careful attention is paid) will have an inherent mismatch. The effect is to convert a small part of the CM currents and voltages into DM noise at the inputs. A differential-mode filter will attenuate interference, which appears between its signal terminals. It will have no effect on interference, which appears in common-mode between these terminals and ground, since there is no parallel capacitance to ground.
The CM filter will attenuate interference appearing between the signal terminals together, and ground. It may also have a lesser effect on differential mode interference. Since imperfectly canceled CM noise can cause CM to DM conversion, it is best to attack the CM interference first, then address the DM interference. Two common types of RFI filters are employed: RC and LC.
RC filtering: op amp configured differentially
Figure 12 shows an example of the use of resistive-capacitive networks for use in an op amp front-end. Because of their sensitivity to input capacitance, RFI filtering is a somewhat trickier process than for in-amps.
The CM filtering is accomplished with R1, R2, C1, and C2. DM filtering is handled by R1, R2, and C3. Since it is the AC CMRR performance that is critical, the time constant of R1C1 should match within 5% of the time constant of R2C2. Also, it is necessary that R3 = R4. At higher frequencies, component matching begins to degrade. To aid in canceling this effect, it is critical to mount each complementary-pair with extreme symmetry with respect to each other (i.e., the same orientation and on the same isothermal lines) over a good ground plane.
The purpose of R3 and R4 is to isolate the amplifier's inverting input from feeling the effects of C3 directly. Without these resistors, the amplifier would have less phase margin.
Figure 12: Filtering differential-mode and common-mode interference in an op amp configured as a differential amplifier. Component matching is essential for maintaining high frequency CMRR.
(Click on image to enlarge)
The feedback time-constant, CFRF, provides first order band-limiting by tailoring the amplifier's closed-loop bandwidth to no more than is necessary to pass the signal without attenuation.
RC filtering with an instrumentation amplifier (in-amp)
Most precision sensor/data acquisition applications use an in-amp as the front-end signal conditioning component, as shown in Figure 13.
Figure 13: Preventing RFI corruption of in-amp signal by low-pass filtering.
(Click on image to enlarge)
Using an in-amp in place of a differential amplifier, we can eliminate six components as well as gain more precision owing to matched input impedances. The combined CM and DM filter is the same, however.
Common-mode -3dB bandwidth depends on the parallel time constants of R1C1 and R2C2, and is:
set to <1/10(amplifier GBW product)
Differential-mode bandwidth is
set between 10× to 40× the highest signal bandwidth of the sensor.
Resistor noise
Because the noise generated by the series resistors used in the filters may be a problem in precision circuits, a careful evaluation of the excess resistor noise should be undertaken.
Resistors contribute noise according to the following equation:
Noise (nV/√Hz) = √4KRT
where: K = Boltzmann's constant (1.38 × 10-23)
R = resistance in ohms
T = temperature in degrees Kelvin (∼300°K at room temperature)
For example, a 5 kΩ resistor (typical for these filters) adds a Johnson noise of 9 nV/√Hz RMS at room temperature to the amplifier's input-referred noise.
The two resistors will produce uncorrelated noise at the inputs, thus the results must be multiplied by √2, which gives a total 13nV/√Hz RMS.
Filtering using common-mode and differential-mode chokes
At the cable end, the RFI encounters a DM choke. Ordinary chokes of sufficient inductance would be un-necessarily large and expensive. The CM choke, by its design, provides an inductance enhancing effect that makes it is far easier to achieve a high impedance than with separate chokes.
As in the RC filter, the most effective way to approach the problem is to address the CM interference first, and then use a separate filter for the DM noise. As shown in Figure 14, the CM choke, in conjunction with two CM capacitors, Ccm1 and Ccm2, is used to suppress the CM interference, producing a large common-mode rejection. Ccm1 and Ccm2 divert any residual CM noise currents to ground and away from the amplifier.
Because even the best CM chokes create some DM currents (principally from leakage inductance) two differential mode chokes, followed by a capacitor across the input terminal of the amplifier, should be added following the CM choke. The two CM capacitors should be grounded to the enclosure, or to the analog ground.
As far as insertion loss is concerned, look for at least 40 dB to 60 dB. More than this will require multi-pole sections. Remember, the filter is to augment the shielding, not take its place.
Figure 14: Using chokes
(Click on image to enlarge)
The EMI/RFI-hardened op amp
Both Analog Devices Inc. and National Semiconductor Corp. have recently introduced new products that address the problem of extracting usable sensor signals in high EMI/RFI environments. The AD8556, and the LMV851, LMV852, LMV854 (single, dual, quad, respectively) series approach the problem of limiting the out-of-band RFI-induced output Vos shifts, common to ordinary amplifiers in the presence of heavy out-of-band RFI at their inputs, in different ways. The ADI device is a digitally programmable sensor signal amplifier that employs an on-chip combination of common-mode and differential-mode filters, in a manner similar to those in our previous discussion. The National parts are high-gain wide bandwidth EMI-hardened operational amplifiers that use proprietary circuit techniques to achieve the same goal.
Uniquely, however, National has a datasheet guarantee of a minimum level of susceptibility over a wide range of frequencies and a means of comparing any op amp with any other. National has coined a term similar to the ubiquitous CMRR found in all op amp datasheets, to quantify the amount of EMI robustness exhibited by each product. The parameter is called EMIRR, or Electromagnetic Interference, Rejection Ratio. It is defined in way analogous to CMRR in that it measures, in real terms, the amount of input referred offset voltage shift that takes place for a given level of radio frequency interference. In this way, an unambiguous comparison of EMI performance can be made between any amplifier and allows a ranking system on the basis of their robustness.
The definition for EMIRR is given by:
Where VRF PEAK is the amplitude of the applied un-modulated RF signal and ΔVOS is the resulting input referred offset voltage shift (see National's application noteAN-1698, "A Specification for EMI Hardened Operational Amplifiers" for a detailed explanation of testing all of the LMV851s, or any other op amp's pins for EMIRR).
Figure 15 shows the EMIRR performance of the LMV85x op amps as a function of input peak RF noise.
Figure 15: EMIRR vs. peak RF input voltage
(Click on image to enlarge)
As described in National's application note, AN-1698, two examples will show how the interfering signals can affect an ADC in the sensor/instrumentation signal path. In both examples an ordinary op amp will be compared to an LMV851. First, suppose in a sensor application, within a hostile RFI environment, that the sensor cable picks up an interfering signal of 900 MHz, which arrives at the input pins of an op amp that has been configured with a gain of ×101. Assume that the RF noise at the input of the amplifier is -20 dBVP and that the amplifier can drive a 10-bit ADC to its full scale input range, which in this case is 5 V. A standard op amp used in this application can easily have a 50 dB EMIRR at 900 MHz. Manipulating the above equation will gives us an offset voltage shift of 0.32 mV. When this is gained up, the output of the op amp is 32 mV into the ADC.
Next, an EMI-hardened LMV851 is placed in the same application. The 80 dB of the LMV851 results in an input referred offset voltage shift of only 10 μV at the 900 MHz. This is equivalent to a 1mV shift at the output. The least significant bit of the 10-bit ADC is 5/1024 = 4.88 mV. The error voltage should always be less than ½LSB, so in this case, it should be no larger than 2.44 mV. The standard op amp had a shift of 32 mV which is equivalent to a 7-bit ADC, while the EMI-hardened LMV851, with its 1 mV of shift, equals 0.2 bit.
Next, Figure 16 compares the EMI robustness of an ordinary op amp to the LMV852 in a pressure sensor application.
Figure 16: Pressure sensor application
(Click on image to enlarge)
In this experiment, the signal conditioner requires two amplifiers configured as a two stage in-amp. Thus, the two op amps are duals. A cell phone is positioned a few centimeters away from the op amps and their signal wires. When a call is made to the cell phone, the op amp and the signal wires are immersed in a strong EM field. The field strength of a cell phone is typically 3V/m and much stronger at a few centimeters. The effect on output voltage of both the standard op amp and an LMV852 is depicted in Figure 17. The typical standard op amp shows a Vos shift more than 1 Vpeak, while the LMV852 is not significantly affected, as a result of the RF signal from the cell phone. The difference in the amplitude of the disturbance for each op amp is equal to the difference in their measured EMIRR.
Figure 15: Comparing EMI robustness
(Click on image to enlarge)
Clearly, the use of an EMI robust amplifier will, in many sensor applications provide a clean output voltage from a sensor signal corrupted by RFI that, heretofore, required expensive filter components and extra real estate on the PCB. However, even with an EMI-hardened amplifier, it is still good practice to use shielded twisted-pair signal cabling. Get the interference as low as possible before it gets to the front-end, and often you will find that no further input filtering is necessary.
About the author
Jerry Freeman is an amplifier applications engineer at National Semiconductor Corp., Santa Clara, CA. He received a BSEE from Heald College of Engineering in 1961.
Editor's note: If this article was of interest to you, also check out:
"Understanding noise optimization in sensor signal-conditioning circuits (Part 1a of 4 parts)",
by Reza Moghimi, click here; note that Parts 1b, 2a, and 2b are linked.