Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

High Frequency Servo shielding and grounding on robot?

  1. Mar 12, 2015 #1
    I have an issue that appears to be relating to noise mitigation and grounding. Some background on myself is I have a mechanical engineering technology degree. But my job requires an ability to work with and troubleshoot electrical issues.

    Currently a customer is using a Servo motor on an end of arm tool for a 6-axis robot. This cell and all of its duplicates run perfectly fine in the build shops. Then run fine for a period of time once initially installed but begin have noise faults on the drive for the servo.

    So the known differences between the build shop and the customer sites are obviously supply power, there is also now a metal detector in close proximity. Then various other noisy machines around it. Various additional grounding has been added to attempt to get the noise away from the tooling and signal lines.

    Something that confuses me about the whole thing is that the low voltage analog signal shield is tied to both ends of the cable?? Meaning it has continuity to the ground and shield of the 480VAC power connector and cable? It also has continuity to the robot arm and robot base ground.

    This goes against all prior experience with robot drives and servos this to my understanding level, is encouraging current to flow across low voltage signal shield. It is also now tying what should be my "clean" grond to my "dirty" ground.

    We have also attached various additional ground straps and this has varying degrees of some cells it helps others it hurts. I am at a loss, does this make any sense or have I provide not enough information? Any input or help would be greatly appreciated.
  2. jcsd
  3. Mar 13, 2015 #2
    This is not my area, but since no one is answering, I'll mention that on one job I worked on, we had to make sure that the signal cable shields were only grounded at one end for the very reasons you mention.

    Can you put up grounded metal screens between the noise and the robot, like a partial faraday cage?
  4. Mar 15, 2015 #3

    jim hardy

    User Avatar
    Science Advisor
    Gold Member

    wow shielding troubles can be tough.

    Some references say that to shield against RF current the shield should be earthed at intervals less than 1/4 wavelength for frequency of interest. What frequency does that metal detector use?

    But earthing a shield at both ends is an invitation for power line frequency currents to flow in it.
    So one is often pushed to find by trial and error what works.

    In my power plant, most of the shields were earthed where they entered the analog equipment rack and floated at the far end.
    That worked pretty well on high level signals of a few volts.

    For microvolt level signals to the plant computer , like thermocouples, we preferred to earth both the thermocouple and its shield at the sensing point - the pipe where it was mounted. Our computer had a floating and guarded analog input system with switched guard, so the shield made a Faraday cage around the whole measurement system. That was an excellent arrangement.

    The basic premise is to minimize capacitively coupled currents into the signal wires.
    To do that you want the shield very near the same potential as the signal wires it encloses, and you don't want much current flowing in the shield.

    So, first thing i'd try is single point earthing of my shields. See which end works better.
    If your troubles are from RF interference not powerline frequency it's a little tougher.

    Here's a short paper

    that agrees with what i read decades ago in this book

    good luck to you, sir - keep us posted ?

    old jim
  5. Mar 15, 2015 #4
    1. Look for anything that may arc, like motor brushes, and relays. brush arcing can be suppressed with line filters grounded with short, wide conductors to the motor body. Relays respond well to these products:
    Or you can use zero-cross relays.

    2. Brushless motor drives are notorious. There's a good method for each configuration (i.e. a motor that runs on isolated electricity versus being tied directly to the AC power.) As a quick test, shield the power leads between the motor and drive and terminate the shield close to motor case, and drive chassis ground. Keep the termination to the shield very short and heavy! A one inch, 18 gauge lead between housing and shield typically does nothing. A 1/4" wide copper braid can usually get the job done up to a couple of inches.
    If the motor drive runs from an isolated supply, then you can terminate the shield to ground for better effect. In a worst case where the motor runs from the AC mains, a three phase common mode filter WITHOUT CAPACITORS can be utilized where the motor drive leads exit the drive.

    Best of luck. Don't hesitate to hire an EMC expert immediately. 95%+ EE's cannot solve these problems, and this machine will likely need more stringent safeguards than making it functional. For example, it will likely need to be EMC and EC tested prior to sale. A 10yr old complaining that your machine interferes with his FM radio is clear grounds to bring a federal lawsuit against you and have your products impounded.
  6. Mar 15, 2015 #5

    jim hardy

    User Avatar
    Science Advisor
    Gold Member

    Thanks Mike
    i forgot that empem's system is probably computer controlled.

    Indeed an arcing relay contact can disturb a differential line computer data bus. That can corrupt data or worse, corrupt an address.
    When your computer queries one location for data but gets back data from a different location because the address got clobbered, the program is going to do something strange.

    We had that happen to us.

    connect with a good software person to characterize these errors.
    We found our trouble only by associating errors with a particular hardware address
    then physically looking at what devices that address served. We saw the sparks and with a logic analyzer were able to find corrupted addresses out in the I/O system whenever sparking occurred. Lots of midnight oil invested in that one.

    The Quencharcs that Mike suggests were our fix and worked exceedingly well.

    Had another case of address corruption caused by a batch of weak line driver IC's from semiconductor manufacturer.
    When they got warm their drive capacity weakened and depending on the strength of reflections on those long data line pairs they might delay a few nanoseconds when asserting a line.
    A 50 nanosecond delay once every twenty minutes brought the system to its knees.
    That one we found by narrowing down the address range, asserting an address in that range 10k times per second and 'scoping all thirty-something lines with an analog scope. Weak drivers showed a bright waveform that was correct with a ghost-like delayed waveform segment stringing out behind so dim one had to turn out the room lights to see it. The bright wave was the 99.9% good assertions, the dim wave was the one in a thousand late assertions.

    I hope your troubles are less evasive than those

    but i mention them to make the point some hardware troubles require the help of a good software guy to encircle them. No shame in that.

    If you cant fix this with shielding, ask a software guy if he can help you associate your errors with some event in the program or with some hardware addresses.
    Hopefully it won't come to that.

    old jim
  7. Mar 16, 2015 #6
    Thanks for all the input everyone! Essentially where my issue comes in is my company just made a j-box of sorts for the signal and motor power on the robot arm. So a lot of the recommendations are well out of the scope of things I have experience.

    Again we have been making similar products for 20+ years and the encoder shield tied at both ends is baffling to us. But the servo company insists it has to be done. Then the cables are overmoded so it doesnt make it easy to do a field test of removing the shield connection at the robot end. Our experience is that robots are noisy mean beings that due to PWM have the ability to induce all ranges of frequency. Thus it is not agood idea to connect the shield on low voltage signal lines at both ends.

    One thing that I have found in the servo documentation I might throw back in their court talks of required use of Ferrite sleeves on instances of long cable runs which this is. As well as line filters.

    At this point the customer is screaming redesign your stuff, but we have no data showing where the noise is being induced exactly and what is really causing it to go bonkers. The analogy I instantly think of is me takingmy car and telling the dealer my car is making a noise. Give me the part to fix it. Dealer asks where the noise is coming from. I tell them I turn left and it stops. There is some help but far to many variables. They do not understand the process of troubleshooting. I am trying to talk them into letting our electrical guru come out to a site with all of his toys. Something I have learned in a production environment. You never talk about religion, politics, or grounding. :)
  8. Mar 16, 2015 #7
    Probably the most common shielding error is terminating at both ends, creating ground loops. Standard practice is the shield should be connected on one and, preferably the end with the source of energy - or closest to the source.

    The encoder shield definatly seems wrong, esp since these are typically run with the Servo Power cable. -

    Also -- Power, Digital and Analog signals should have separate shielding / grounding networks - the robot can easily have all three running to the arm.
  9. Mar 16, 2015 #8
    Windadct, Your last statement, this is what confuses us with this servo also. In our products we insulate and isolate the power and signal wiring and connectors. But the servo drive manufacturer is telling us in our block we have to tie the power shields and signal shields together. Stating by isolating them we are creating an impedance...

    One thing I requested was a large braided ground strap from the tooling (connectoed to the frame of the servo motor) all the way back to the ground plain. This should offer a nice low impedance bond I would think? It wouldnt last long on a robot arm due to being braided, but would be a good test.
  10. Mar 16, 2015 #9
    Not knowing what the tooling is- Something like a plasma, welder or even waterjet - can develop static or basic electrical noise, a ServoMotor - I guess could have some EMI due to the drive signals hitting the impedance of the motor, the motor housing should be grounded, but it should not need to be particularly large.. I works on a lot of these, and we typically had separate cables for Power and signals for the larger motors. I am assuming these are all BLDC servo motors, correct ( just to confirm there are no motor brushes involved)

    If one system is particularly troublesome, can you temporarily lift the shield(s) on the side on the robot? - so they are all terminated on one end only(controller side typically) -- I have never seen anyone advise to connect both ends of a shield - this loop is then run with the power cables? --
  11. Mar 16, 2015 #10

    jim hardy

    User Avatar
    Science Advisor
    Gold Member

    never talk grounding - i like that.

    There's grounding and there's shielding...

    Here the safety shield should be "grounded" to earth
    but not the signal shield. The signal shield goes to circuit common which may or may not be earthed. Observe it makes a Faraday cage surrounding the whole measurement system.
    And the signal shield will be terminated at both ends but not necessarily to earth ground.


    That's why i dislike the term "Ground", preferring instead "Signal Common" and "Earth" to eliminate ambiguity.

    drawing is from a good article at
  12. Mar 16, 2015 #11

    jim hardy

    User Avatar
    Science Advisor
    Gold Member

    "overmoded" ???

    I tried looking it up and got links to cavity resonance phenomena.

    Are you guys really that much ahead of me? No surprise if so, but would you point me to an introduction ?

    old jim
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook