4-20mA vs 0-5V ratiometric output; how much noise rejection do I really need?

Thread Starter

ebeowulf17

Joined Aug 12, 2014
3,307
I'm looking for general advice on how to know (up-front) when you need the benefits of a 4-20mA output. I'm working on electrical design and sensor choices for an updated commercial espresso machine design. I know from past experience that there are numerous noise sources in our machines that cause various problems:
  • Arcing relays for 1/3HP pump motors have caused RS485 communication dropouts fairly often, sometimes fouled the data for a 4x20 LCD display, and even caused microcontroller reboots occasionally.
  • High current wires for heating elements (over 10A) induce noise in nearby RTD sensor wires, skewing temperature readings (admittedly, a better wiring layout would help a lot with this.)
  • Numerous solenoid valves all create smaller noise spikes throughout the machine, occasionally causing problems similar to the motor problems described above (local snubbers help, but aren't 100% reliable.)
Ideally, when we make updates or create whole new designs, we'll get all the little details right and eliminate most of these noise issues at their source. Nevertheless, past experience tells me that even when we try to do it right, we can't always prevent all the noise, so I've wanted to do whatever we can to make other circuits immune to this noise as much as possible. We're currently looking at adding pressure transducers to the new design (datasheets for two of the possible candidates are attached below) and I thought that I wanted to go with 4-20mA current loop output for it's greater noise immunity. However, the Honeywell model, which otherwise appears better in several ways, doesn't offer that option (the Sensata part doesn't list it, but we've heard from Sensata that they can make that version for us.)

So, I'm trying to decide how much I should prioritize 4-20mA output. Is it a deal-breaker? We're dealing with pretty short wire runs - possibly only a foot or two if we do what I want, and less than 10 feet long even in the worst-case scenario. So, the voltage drops over long wire runs aren't an issue. I'm just wondering about noise on a low-impedance vs high-impedance signal carrier. I know that inputs with really high impedance, like multimeters, oscilloscopes, op amps, etc. can be incredibly sensitive to any sort of induced noise, and that current loop inputs are able to shed most of that because they're low impedance. What I don't know is where the cut-off is.

The Sensata part (0.5 - 4.5V ratiometric output) recommends a 10kΩ pull down as the load (input impedance on the board we would design.) The Honeywell part recommends a load >2kΩ. Are these input impedances low enough to help with noise rejection, or would 4-20mA current loop be a much safer bet?

For what it's worth, response time needs to be somewhat quick as it's part of a closed-feedback loop, but I'm sure a few milliseconds of lag wouldn't hurt anything, so we can add some amount of filtering too, whether in the form of input capacitors, digital filters, or both.
 

Attachments

TeeKay6

Joined Apr 20, 2019
573
I'm looking for general advice on how to know (up-front) when you need the benefits of a 4-20mA output. I'm working on electrical design and sensor choices for an updated commercial espresso machine design. I know from past experience that there are numerous noise sources in our machines that cause various problems:
  • Arcing relays for 1/3HP pump motors have caused RS485 communication dropouts fairly often, sometimes fouled the data for a 4x20 LCD display, and even caused microcontroller reboots occasionally.
  • High current wires for heating elements (over 10A) induce noise in nearby RTD sensor wires, skewing temperature readings (admittedly, a better wiring layout would help a lot with this.)
  • Numerous solenoid valves all create smaller noise spikes throughout the machine, occasionally causing problems similar to the motor problems described above (local snubbers help, but aren't 100% reliable.)
Ideally, when we make updates or create whole new designs, we'll get all the little details right and eliminate most of these noise issues at their source. Nevertheless, past experience tells me that even when we try to do it right, we can't always prevent all the noise, so I've wanted to do whatever we can to make other circuits immune to this noise as much as possible. We're currently looking at adding pressure transducers to the new design (datasheets for two of the possible candidates are attached below) and I thought that I wanted to go with 4-20mA current loop output for it's greater noise immunity. However, the Honeywell model, which otherwise appears better in several ways, doesn't offer that option (the Sensata part doesn't list it, but we've heard from Sensata that they can make that version for us.)

So, I'm trying to decide how much I should prioritize 4-20mA output. Is it a deal-breaker? We're dealing with pretty short wire runs - possibly only a foot or two if we do what I want, and less than 10 feet long even in the worst-case scenario. So, the voltage drops over long wire runs aren't an issue. I'm just wondering about noise on a low-impedance vs high-impedance signal carrier. I know that inputs with really high impedance, like multimeters, oscilloscopes, op amps, etc. can be incredibly sensitive to any sort of induced noise, and that current loop inputs are able to shed most of that because they're low impedance. What I don't know is where the cut-off is.

The Sensata part (0.5 - 4.5V ratiometric output) recommends a 10kΩ pull down as the load (input impedance on the board we would design.) The Honeywell part recommends a load >2kΩ. Are these input impedances low enough to help with noise rejection, or would 4-20mA current loop be a much safer bet?

For what it's worth, response time needs to be somewhat quick as it's part of a closed-feedback loop, but I'm sure a few milliseconds of lag wouldn't hurt anything, so we can add some amount of filtering too, whether in the form of input capacitors, digital filters, or both.
A very simplified answer is that if the signal is being corrupted while in transit from source to destination, then a current loop may help. If the problem is that the source signal is corrupted prior to transmission or the received signal is corrupted after receipt, then the current loop is less likely to be of value. Thus, the first step is to determine what kind of noise you have and where it originates along the transmission path. All current loop schemes have limits for the allowed range of voltage of both wires of the loop; your design must consider those limits at both the source and destination. If a corrupting signal is able to cause the voltage of either wire to exceed its limit, then the loop accuracy suffers. Using a twisted pair, optionally with an overall shield, helps to ensure that corrupting influences affect both wires equally. A current loop is most commonly used to eliminate "ground loops" (differences in ground potential between the transmitting and receiving ends of the loop), typically at power line frequencies. There are limits to what frequencies can be eliminated; for example, if a high frequency wire runs parallel to one of the current loop wires, then coupling via stray capacitance can alter the voltage of the loop wire and that effect is worse as frequency increases. A full answer could fill a book. If your knowledge of current loops is limited, then something like this might help: https://instrumentationtools.com/why-use-a-current-loop/
 

Thread Starter

ebeowulf17

Joined Aug 12, 2014
3,307
A very simplified answer is that if the signal is being corrupted while in transit from source to destination, then a current loop may help. If the problem is that the source signal is corrupted prior to transmission or the received signal is corrupted after receipt, then the current loop is less likely to be of value. Thus, the first step is to determine what kind of noise you have and where it originates along the transmission path. All current loop schemes have limits for the allowed range of voltage of both wires of the loop; your design must consider those limits at both the source and destination. If a corrupting signal is able to cause the voltage of either wire to exceed its limit, then the loop accuracy suffers. Using a twisted pair, optionally with an overall shield, helps to ensure that corrupting influences affect both wires equally. A current loop is most commonly used to eliminate "ground loops" (differences in ground potential between the transmitting and receiving ends of the loop), typically at power line frequencies. There are limits to what frequencies can be eliminated; for example, if a high frequency wire runs parallel to one of the current loop wires, then coupling via stray capacitance can alter the voltage of the loop wire and that effect is worse as frequency increases. A full answer could fill a book. If your knowledge of current loops is limited, then something like this might help: https://instrumentationtools.com/why-use-a-current-loop/
Thanks for your thoughts, and for the link. I read up on current loops several years ago, but I've only dabbled in them in simple scenarios with short short wire lengths. I made a circuit to power and read the output from a pressure sensor sender, but that's about it.

I read your linked article, and I've read several others recently, and the impression I'm getting now (which is different than what stuck in my memory from a few years ago) is that the biggest benefit is around voltage drops on long wire runs, which isn't an issue for me. All of the articles also mention noise rejection in passing, but they almost make it sound secondary to wire length and voltage drop benefits.

On the other side of things, how high of impedance is too high? If I use one of the ratiometric 0.5-4.5V output sensors, how bad is the required 2k to 10k input impedance in terms of noise? I know really high input impedances in the megaohm range practically turn wires into antennas, but how much better is 10k? Where do people generally draw the line between "high impedance" and "low impedance" inputs as far as noise rejection?
 

TeeKay6

Joined Apr 20, 2019
573
Thanks for your thoughts, and for the link. I read up on current loops several years ago, but I've only dabbled in them in simple scenarios with short short wire lengths. I made a circuit to power and read the output from a pressure sensor sender, but that's about it.

I read your linked article, and I've read several others recently, and the impression I'm getting now (which is different than what stuck in my memory from a few years ago) is that the biggest benefit is around voltage drops on long wire runs, which isn't an issue for me. All of the articles also mention noise rejection in passing, but they almost make it sound secondary to wire length and voltage drop benefits.

On the other side of things, how high of impedance is too high? If I use one of the ratiometric 0.5-4.5V output sensors, how bad is the required 2k to 10k input impedance in terms of noise? I know really high input impedances in the megaohm range practically turn wires into antennas, but how much better is 10k? Where do people generally draw the line between "high impedance" and "low impedance" inputs as far as noise rejection?
First, I see in the Honeywell MIP series datasheet that that series (and perhaps others) is ratiometric, as you claim also is the Sensata sensor.

Searching the Sensata website I can find only one pressure sensor series that claims to be ratiometric, the 60CP series. Nowhere in the website can I find an indication that the 2CP5/2CP50 is ratiometric. Otherwise, the two series do have very similar specs. Unfortunately, Sensata provides very meager technical info on either series and no application literature. My interpretation, based on using ratiometric design of other circuitry, is that a ratio metric sensor would act very much like a potentiometer, with the input DC excitation (5V for Sensata) being divided proportionally to pressure and appearing at the wiper of the potentiometer. In reality the circuitry is more complex, with the output signal almost certainly buffered by an opamp or similar electronics. Thus, in use, a ratiometric output is a ratio of the excitation voltage. For example, at 50% of the F.S. pressure range, the output would be 50% of the excitation voltage. What this means to me is that the same excitation voltage must be available both at the sensor and at the point of use of the output--for best accuracy. The user's circuitry must properly interpret the output voltage as a fraction of the excitation voltage. The sensor output could be fed directly to an analog-to-digital converter (ADC) whose reference was the excitation voltage; the ADC would then give stable & accurate output despite variations in the excitation voltage. This is, I believe, of no value to you as you intend to use the sensor output directly as an analog signal. Here are some links to info on ratiometric pressure sensors:

https://www.sensorsone.com/ratiometric-0-5-to-4-5vdc-pressure-transducers/
https://appmeas.co.uk/resources/pressure-measurement-notes/what-is-ratiometric-output/
https://www.carel.com/sensor-pressu..._content/56_INSTANCE_i4q5KIMLInKK/10191/46930
https://www.globalspec.com/industrial-directory/ratiometric_pressure_sensors

To be clear, the output of the ratiometric sensor is still an analog output. However, the output will vary as the excitation voltage varies (for a fixed pressure). Thus, to obtain a stable analog pressure reading the excitation voltage must be stable (vs temperature, time, aging, etc). Assuming this okay for you, then let's discuss resistance values.

First, there is no dividing line between low and high impedance nodes; what is low in one instance might be considered high in another instance. Essentially, in the context of your question, a high impedance node would be a node whose voltage is susceptible to corruption by nearby noise (all electrical kinds) to an unacceptable degree for that specific application. That is, if you are trying to discern changes of 1uV in the pressure signal, then corruptions of 1uV would be intolerable; however, if you were attempting to discern changes of 1mV, then corruptions of 1uV would like be okay. It it nevertheless true (I am sure that there must be some exceptions) that a lower resistance/impedance node is less susceptible than a higher resistance node. A useful spec, not given by Sensata or Honeywell, would be the output impedance of the sensor; that is, I believe, more relevant than their spec for load resistance.

Now we are back to my earlier statement. If you want to rationally eliminate noise, then you must first find its source and then take appropriate design steps to reduce the effect of the noise. Unless you know what you are fighting, you don't know what weapons to use.
 
Last edited:
Top