Several single ended output and PSU powered transducers are tens of meters apart from each other outputting 0 to 10V DC like voltages.

Some are temperature transducers some are force transducers ect. Their outputs all go to a single DAQ board which essentially multiplexes these channels and logs the data.

The DAQ board is also single ended and non referenced earth grounded and named as NRSE in the below illustration. The transducer signals are conveyed through BNC coaxial cables. Tr represents transducers below:

enter image description here

I’m planning to star-ground all power supply negative terminals and sensor shields at a single point and earth ground them. So none of the sources will be floating wrt earth. Since the DAQ is NRSE its analog ground will be floating with respect to earth.

I only found this document related topic http://www.ni.com/white-paper/3344/en/.

And mine is the one below encircled in red:

enter image description here

In my case, as I have explained with my first illustration many inputs actually nine inputs from transducers apart from each other will be coupled to the DAQ in NRSE configuration.

Do I really need to star-ground the DC negative terminals and transducer shields at a single point? SG is the signal Ret is the return wires going to the AISENSE. (For a particular case I got a better result earthing the DC terminal instead of return wire; it was CM interference due to SMPS transformer), thats why I earth ground the transducer’s from their power supply negative terminal in my illustration.


Basically what Im asking is the star connected node(red question mark) need to be earth grounded as well?:

enter image description here

Edit: enter image description here

AI SENSE is like diff ended with common returns.

AI GND and AI SENSE should not be connected thats what I heard from manufacturer for AI SENSE used scheme. Figure 1 SE scheme must not be used it causes parasitically coupled ground loops.

Now the transducers are powered by power supplies and power supply ground and transducer ground is not isolated.

In Figure 2 if I use AI SENSE and if I also route PSU ground to AIGND or to the nearest earth I get very good results. So it seems to me if I use AI SENSE scheme and a third wire from the red point(the PSU ground) to the nearest earth(yellow) or to AIGND(green) things are fine.

But since I have many of these I was asking should I use the green or yellow route in Figure 2.? Also there are 10 inputs in that case where should these yellow or green wires be star grounded?

But I can assure you SE scheme and tying AI GND to AI SENSE cannot be done. So please the answer be none of that two. AI GND is always earthed through PC motherboard. AI SENSE should not be wired to AI GND. Pls provide a simple drawing in the answer makes things easier to follow.

  • \$\begingroup\$ I think this depends on the existence of electric fields that couple onto the wiring from the sensors. Those electric fields will induce charges, with these charges taking all possible paths back home. Will you have cleaner measurements, if YOU provide the lowest path (the earth ground)? By the way, I upvoted your question, because I want to see how experts (I ain't an expert) will handle this question. \$\endgroup\$ Commented Oct 23, 2018 at 5:23
  • \$\begingroup\$ Please update your diagrams with what you plan on doing with the DAQ ground, and whether the power supplies will be located close to the DAQ or with the sensors. Pleas also show what the DAQ is and analog and whether there is any isolation \$\endgroup\$
    – Voltage Spike
    Commented Oct 25, 2018 at 18:00
  • \$\begingroup\$ To answer this question it's important to know how TR1..3 work. Is their output isolated from the input? Can you add schematics of their output stage? If it's not you have your first ground loop in your first image. \$\endgroup\$
    – Andy
    Commented Oct 27, 2018 at 8:19
  • \$\begingroup\$ Made an edit hopefully more clear. \$\endgroup\$
    – floppy380
    Commented Oct 27, 2018 at 16:53
  • \$\begingroup\$ Just a comment; not an answer: with cables this long you may have substantial noise problems. A common solution is to use transducers with current output (4-20mA) instead of voltage output (0-10V). I suppose it's too late for this project, but something to think about... \$\endgroup\$
    – bitsmack
    Commented Oct 28, 2018 at 8:17

3 Answers 3


First off, system designs like this are an art not a science. It is an art because there are many parameters of the system design that are unknowable (or we don't have time to find them). One could go through the whole system and model it as a giant circuit. To do so one would have to measure/estimate the cables and the noise sources (RFI and conducted emissions through mains AC) takes more time than using best practices.

Short answer:

Basically what I'm asking is the star connected node(red question mark) need to be earth grounded as well?

No, a star ground is only going to add more ground loops and more noise, since you are likely in a good configuration already as per this comment:

In Figure 2 if I use AI SENSE and if I also route PSU ground to AIGND or to the nearest earth I get very good results. So it seems to me if I use AI SENSE scheme and a third wire from the red point(the PSU ground) to the nearest earth(yellow) or to AIGND(green) things are fine.

The idea is to do what works for the noise levels the system needs to attain. If you build the system, and the noise is acceptable for the measurements that need to be made, then your done! Figure 1 would be preferable to figure 2, remember the figures in the NI document are best practices. The environment or setup that you have might not have as much noise, really you would need to have a diagram of the power supply, how its connected to each transducer and a knowledge of the grounding of the DAQ.

Noise and Ground Loops

If the system does have too much noise, then take steps to eliminate it. I'll layout some of the best practices in the event that you do experience noise:

The first concept to understand is that of a ground loop. The ground loop starts because there are two points of grounding with two cables. This forms a large loop, if magnetic fields run through the loop (from motors and power lines) it creates a current that runs around the edge of the loop. This creates a problem even with shielded cables via inductive coupling (shielded cables only protect against capacitvely coupled electric fields and block them, but not magnetic), in which a current on the outside of the cable also causes a current on the inside. In most cases it is best to avoid ground loops.

With a DAQ this creates problems because you might not have total control of the grounding of the sensor wires or the shield.

If ground loop currents are the biggest source of noise (which depends on the electric\magnetic environment that the system is in) then the best thing to do is break the loop. This can be done by removing one ground (best practice) or by disconnecting the shield (preferably next to the sensor.

Another thing that can be done is to add isolation to the analog signals to break a ground loop, this can be more expensive, but is better.

enter image description here
Source: http://www.sensorland.com/HowPics/IPC07-004.gif
Or this page

It is also good to use twisted pair wire to eliminate loops and keep noise down (this is why most differential signals like RS485 or ethernet have twisted pair.

Good practices

If you ever build one of these systems again, it would be more convenient from a noise perspective to put the power supplies in the same location as the DAQ. Having a ground on one end is usually the best condition, sometimes the cable can turn into a radiator (antenna), in that case a ferrite can block high frequency signals on the cable. It would also be good to provide shielding around the sensor/source/amp on the end of the cable and tie it to the shield (if the shielding is not grounded, if it is you can break the shield.)

enter image description here

Watch grounding on both ends as this generates currents if you can help it.

enter image description here

It's not a good idea to put a power supply on the opposite end unless its isolated.

enter image description here


There is no reason to have your analog inputs Earth grounded. It does nothing except offer a path for noise to get into your analog signal.

I have assembled some complex ATE equipment with all types of issues. First rule is each source stands on its own until it reaches a differential op-amp or ADC MUX.

Get rid of all your red and green lines. The black signal wires are ok as is. Return all blue wires to the AIsense pin or use a dual-port MUX. Connect it to common signal ground if the PGIA datasheet shows it to be necessary.

You want to avoid having AIsense cross-connected before reaching the PGIA. This avoids signals leaking into other channels.

Do NOT connect high energy devices directly to these inputs or AIsense. It can damage the PGIA and MUX and ADC. Use an analog isolation module (AD210KN) for such devices.

Avoid earth grounds, it's just a long wire that acts as a source for RF noise pickup. Keep what wires you need short and twisted if you cannot shield them using STP (Shielded Twisted Pairs). If you do use shielded wires the ground of the shield should only connect to signal ground at the PGIA or close to it. The shield wire must be quiet and that is a quiet spot.

In the below schematic a dual 4:1 mux allows the signal to select only 1 sensor at a time. Channel 'B' is not used. Since the sensors always have an AI sense connection they do not 'float'. The CD4052 MUX is very low cost and has high channel to channel isolation. Read the data sheet before using it.

There is NO connection to Earth ground and signal sources never share a common connection until they reach the AI sense pin. That is your 'star' point. If not selected the sources have no load, but that should not be an issue for a source. If sensor wires have a long run then add clamp diodes to Vss and Vee (1N4148) which will protect the CD4052 from voltage spikes over 10 volts, but sensor wires should not run anywhere near AC power feeds.

Note that figure 4 in your NI documents is almost identical to this set up except you have just a 4 channel SE input instead of 8.


simulate this circuit – Schematic created using CircuitLab

Link to CD4052 datasheet

  • \$\begingroup\$ The mafufacturer says use AI SENSE but do not connect AI SENSE to AI GND. \$\endgroup\$
    – floppy380
    Commented Oct 30, 2018 at 17:59
  • \$\begingroup\$ Please add a simple drawing if possible \$\endgroup\$
    – floppy380
    Commented Oct 30, 2018 at 17:59
  • \$\begingroup\$ +1, with one caveat: The common mode voltage of the input signal with respect to the DAQ will not be controlled as it would be if the signal were ground referenced. So it's possible that the common mode voltage at one end could diverge from the common mode voltage at the other end to an objectionable level, and that could result in damage to one end or the other. This is more of a problem in widely distributed systems with signalling across power domains, but in all cases the risk should be understood and mitigated through added protection or careful wiring practices. \$\endgroup\$
    – ajb
    Commented Oct 30, 2018 at 19:37
  • \$\begingroup\$ @ajb. I understand these issues, Even a high-impedance ground may help with power supply leakage noise. OP seems to be trying to get this right based on calls to tech support. They recommend that AIsense not be tied to signal gnd. AIsense maybe a bias voltage and not just a common return. \$\endgroup\$
    – user105652
    Commented Oct 30, 2018 at 19:42
  • \$\begingroup\$ @user1234. I changed the schematic so you now have SE inputs with a 'star' topology for the AI sense return, at the AI sense pin of the PGIA. I removed the connection from AI sense to signal ground to avoid confusion. I modified my answer as well. \$\endgroup\$
    – user105652
    Commented Nov 2, 2018 at 4:52

You need to have the AI SENSE not floating in regard to the PGIA supply, while avoiding to have ground loops.

If AI SENSE is floating, you will have common mode noise which will affect your measurement so you need to tie down AI Sense on the DAQ side.

Also the tie down of AI Sense needs to be done before a MUX, which is the case in your design since you don't MUX the negative side but I just mention this in case.

On your first diagram, you may have ground loops if you tie down AI Sense to the ground:

enter image description here

The best solution in this case, to have no ground loop and to have GND tied to AISense, would be to:

  • Connect the PGIA GND to the AI Sense.
  • If the cables have shielding, connect to PGIA GND only on the DAQ side.
  • Leave the sensors GND of the PSU1,2,3 floating.

In this way you should have a clean reference PGIA GND free of ground loops.

You can improve further by having 3 wire cable with 2 sense line which are twisted (AI Sense and AIx) + 1 GND "power" cable + a shielding.

In this situation, AI lines only carries signal, the GND cable bring everything at the same ground and carries eventual currents, shielding and twisting further reduce induced noise.

Tie the GNDs of the PSU to the GND "power" line and the AI GND, and tie AISense to this GND through a 100 ohm resistor (you may want to try different values). The resistor is there to limit current flowing on the negative sense line.

It's important to connect the shielding only on the side of the DAQ, to avoid ground loops again and avoid having current flowing on the Shielding.

  • \$\begingroup\$ AI SENSE has to be. I cannot ground it and the sources will not be floating. The criteria is clear in the question. Single ended earth grounded with floating sources you suggest is horrible scheme I have suffered a lot from it. No more that. \$\endgroup\$
    – floppy380
    Commented Oct 26, 2018 at 18:24
  • \$\begingroup\$ If the source cannot be floating, you will have to get the a galvanic isolation on each channels to avoid ground loops and that will increase the cost significantly. I can provide a solution for that if you want to do the electronic yourself otherwise look for such device on the market. \$\endgroup\$
    – Damien
    Commented Oct 29, 2018 at 4:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.