Avoiding Noise when Interfacing with Remote Sensors

Mark Harris
|  February 4, 2021
Avoiding Noise when Interfacing with Remote Sensors

This article looks into the problems associated with noise pick-up when interfacing with remote sensors. An excellent place to begin is with a circuit that many of us will be familiar with but universally renowned as being very poor at dealing with dominant noise sources such as electric field interference.

Non-differential op-amp interface

In this example, a remote sensor (unearthed) is connected to a single-ended op-amp via a cable. 

remote sensor (unearthed)

If we add transient interference from an external electric field, we will see in the diagram below that this affects the two wires in the cable differently:

transient interference

Because the two wires have different impedances to ground, the red transient voltages will be different. The transient on the non-inverting input will be smaller because the energy will be conducted directly to ground. The series resistor’s value will determine the transient on the wire feeding the inverting input. Also, the sensor’s output impedance cannot be relied upon; it may equalize the transients to some extent (at its end of the cable) but won’t balance the discrepancies at the op-amp end.

This discrepancy in transient voltages will be amplified by the op-amp (along with the wanted signal), producing a noisy output.

Using a Differential Amplifier

The situation can be significantly improved by using a differential-connected op-amp, as shown in the diagram below:

differential-connected op-amp

Because the impedance to ground on both wires is balanced, the red transient voltages that feed the differential amplifier are the same. This cancels out the noise, and the op-amp output will be clean. If the resistors used are not exactly the same value, there will be a small impedance imbalance. However, the op-amp output noise will still be significantly smaller than when using a non-differential amplifier circuit.

Sensor impedance imbalance

Previously, the sensor has been assumed to be floating, but this will not always be the case. The sensor’s output impedance (X) can become a problem. Consider the following scenario:

The sensor’s output impedance (X) can become a problem.

The two wires in the cable now present different impedances to ground because of the sensor’s output impedance (X) on the upper wire. This means that electric field interference will produce different transient voltages at the differential amplifier’s inputs and lead to noise appearing on the op-amp output.

To counter this problem, we need to “impedance balance” the sensor output with another resistor that matched the sensor’s output impedance (X):

The sensor’s output impedance (X) can become a problem.

The impedance balance is restored, and the op-amp output will again be largely free of noise. This is a balanced impedance transmission system. Whether you are trying to measure analog signals from a sensor or provide an interface for remote digital signals, the basic methodology will be the same.

Adding more immunity

Although the final scenario above paints a rosy picture, one thing that can degrade performance is the op-amp’s common-mode rejection (CMR) performance at higher frequencies. We have been discussing external electric field interference problems, which boil down to the equivalent of connecting a source of interference to both wires via small value capacitors. This means that the interference tends to be greater at higher frequencies due to the lower capacitive reactance. In other words, noise coupling is more significant at high frequencies.

We can see the problem of CMR by looking at the datasheets of two prospective op-amps. For this comparison, I’m looking at the OP07 and the OP1177 devices:

 OP07 and the OP1177 devices:


Both devices are similar when it comes to their offset voltages and errors. Both support approximately the same power supply range; hence, they both could be considered for use in this type of application. However, if we look at the CMR graphs, we start to see why the OP1177 is superior.

At 1 kHz, the OP07 CMR is about 93 dB, whereas, for the OP1177, it’s about 110 dB. For the OP07, 93 dB isn’t too bad a figure, of course; that’s a 45,000:1 rejection of common-mode voltages. However, the problem starts to show itself at higher frequencies. At 10 kHz, the OP07’s CMR is about 73 dB (about 4,500:1), and above about 25 kHz, the graph ends, and we cannot rely on rejection performance at higher frequencies.

The OP1177 can be relied upon up to 10 MHz, with the lowest CMR being 60 dB. That’s a rejection ratio of 1000:1. If we straight-line projected the CMR graph for the OP-07 op-amp up to 10 MHz, we might see only 3 dB rejection at 10 MHz. I’m not saying it’s fair to make this numerical assessment for the OP-07, but how else could we make a comparison between it and the OP1177. If MHz range frequencies dominate the electric field noise, we can either choose an op-amp that can handle this or consider a CM filter.

Common-Mode (CM) filters

Because the OP07 is cheaper than the OPA1177, we might still consider using it if we can add a common-mode (CM) filter at the inputs. The simplest method would be to attach a capacitor from each input wire to ground before the differential amplifier. We also have to be conscious that this may destabilize the sensor output if it uses an internal amplifier so, we will add two more series resistors (Y) to the circuit as shown in the following circuit diagram:

 add a common-mode (CM) filter


Firstly, let’s talk about the downside of adding a CM filter. The tolerance of the added capacitors means that it may degrade CMR. Suppose there are equal CM noise voltages (shown at the red circle points before the filter), then after the filter (across each capacitor). In that case, there may be unequal levels to the tune of 10% if we use 5% tolerance capacitors. That is, one capacitor might be 5% low in value while the other might be 5% high, which results in a differential voltage that may, worst case, be 10% of the CM voltage. This is the downside of adding a CM filter - it creates a differential noise level that could be about -20 dB for 5% capacitors.

The upside is that the filtering will heavily attenuate the CM levels, so we might still get a benefit even if the much smaller levels are not precisely balanced. Therefore, in this example, to get a net CM reduction of 60 dB will require that the low-pass RC filter attenuation is 80 dB. If the resistors (Y) are chosen to be 10 kΩ, then the capacitive reactance would need to be 1 Ω to produce a general 80 dB reduction of the unwanted CM noise.

If we targeted 1 kHz as the frequency where we want this attenuation, this means that the capacitors would need to be 159 𝜇F, and this is clearly too great a value to justify. Finding inexpensive 5% capacitors of this value is also problematic. So that sets the scene for the added resistors (Y) to be in the region of 1000 kΩ. But no matter what resistor values we choose, there is no escaping the fact the CM filter will also bandwidth-limit the desired sensor signal to 0.1 Hz. 

So, in many applications, we’re fighting a losing battle when trying to find a solution based on the lower performance op-amp. There are, having said that, some applications where CM filtering might still be needed for the OP1177 op-amp, but such filtering wouldn’t need to “kick-in” until we reach frequencies above 1 MHz. At 1 MHz with Y = 10 kΩ, the capacitive reactance is 1 Ω, and that means a capacitance of 159 nF. However, the signal bandwidth is still relatively poor at just 100 Hz.

Remember that this is just an example to show the pitfalls of thinking that a CM filter is a panacea for fixing noise problems. They can be tricky to design, and it’s usually cheaper in the long run to choose a better op-amp.

Magnetic field interference

Providing that the interfering current (orange) induces equal voltages into each cable wire, the previous solutions we’ve looked at for electric field interference are equally valid for combating magnetic field interference:

interfering current will induce voltages in both cable wires


The interfering current will induce voltages in both cable wires. Due to the balanced impedance of the whole circuit, the induced voltages will be equal in amplitude at the input to the differential amplifier. You have a loosely coupled transformer with two secondary windings. The primary is the interfering current shown as an orange wire and, if both “secondaries” (the cable wires) are equally electrically loaded (which they are), they will produce the same CM terminal voltage. Hence, the noise will be canceled at the differential amplifier output.

Screens and twists

Previously, we have seen that CM filters can be somewhat tricky to design. However, the most effective method of reducing electric field interference is by using a simple cable screen. The screen causes interference to “dose” both wires more equally, but the more significant advantage is that it acts as a Faraday cage. This results in the interference generating much lower CM voltages and generally means that a differential amplifier will cope better in “harsher” environments. 

The screen, however, does need to be grounded to make the maximum gain. Grounding at the receiving end is the preferred method. Grounding at both ends is a nice idea, but then it becomes a ground-loop that can potentially conduct fault currents from “other sources.” This then usually becomes more of a problem than a solution, but there are some circumstances where low-value capacitors can “ground” the cable screen at multiple points. Their low value provides a high impedance to ground for fault currents but a low impedance for higher frequency noise.

Twisting the two wires in the cable ensures that both wires will roughly be equally distanced from any localized source of interference (electric field or magnetic field). However, given that a cable screen has minimal effect on any prevailing magnetic source of interference, twisting provides a more significant benefit.

Conclusion

This article can’t hope to cover all scenarios, so we’ve just focused on the most common. We have seen that a fully impedance balanced circuit can be the right approach and demonstrated that both electric and magnetic sources of noise could be significantly reduced when using this approach. Simple RC CM filters are shown to be a mixed blessing and probably not cost-effective as a solution compared to spending the extra money on a better performing op-amp. Finally, all the effects that we have discussed can be reproduced using a simulation tool, and I recommend that this is the best approach to take for your own specific problem.

The design tools in Altium Designer® contain everything you need to keep up with new technology. Talk to us today and find out how we can enhance your next PCB Design.

About Author

About Author

Mark Harris is an engineer's engineer, with over 12 years of diverse experience within the electronics industry, varying from aerospace and defense contracts to small product startups, hobbies and everything in between. Before moving to the United Kingdom, Mark was employed by one of the largest research organizations in Canada; every day brought a different project or challenge involving electronics, mechanics, and software. He also publishes the most extensive open source database library of components for Altium Designer called the Celestial Database Library. Mark has an affinity for open-source hardware and software and the innovative problem-solving required for the day-to-day challenges such projects offer. Electronics are passion; watching a product go from an idea to reality and start interacting with the world is a never-ending source of enjoyment. 

You can contact Mark directly at: mark@originalcircuit.com

most recent articles

Back to Home