Ground Plane (PCB) Philosophy

Discussion in 'General Electronics Chat' started by ue418, Jun 22, 2010.

  1. ue418

    Thread Starter New Member

    Jun 22, 2010
    17
    1
    I currently have 4 signals (+/- excitation and two returns) going between point A (where the two returns are amplified) and point B (a loadcell/Wheatstone bridge). I have a small circuit at Point A (mostly containing the instrumentation amp), but my plan is to replace it with a much larger circuit board. Currently, the four signals go from A to B in the form of a shielded cable, with the shielding being drained/grounded on one end (the Point A end). But in my envisioned design, the circuit board would cover most of that distance (about 6 inches) itself, so that the four signals would run along traces in the new PCB. Note that I am shielding these signals because I have a nasty stepper motor sitting nearby. So, my question is...how do I duplicate/simulate the shielded cable when I switch over to a pure board design? My gut says that I should have a multilayer board with these particular traces on one layer, and with ground planes both above it and below it. Of course, these ground "planes" would not have to cover the whole area of the board, I am telling myself. It seems they should follow the four traces, but extend out on either side by a half-inch or so? So they would basically be two inch-wide sheets of copper, following the exact path of the traces sandwiched between them. However, I haven't read much about this sandwiching technique when it comes to PCB design. Do I have this all wrong? Do I only need one ground plane? I could really use some advice, as I may need to build 15 or 20 of these machines quite soon.
     
  2. Ghar

    Active Member

    Mar 8, 2010
    655
    72
    That sandwich technique is called a stripline, you'll have better luck reading about it using that term. The effect will work as long as the planes are much wider than the signal trace, I believe the rule of thumb is a factor of 5.

    If the noise source is the motor have you tried reducing it's noise? Twisting its wires together (power with return) is a good first step.
     
    ue418 likes this.
  3. nanovate

    Distinguished Member

    May 7, 2007
    665
    1
    Can you put the instrument amp on side B so that you are transmitting an amplified signal?
     
  4. timrobbins

    Active Member

    Aug 29, 2009
    318
    16
    How does your present config perform wrt noise ingress? If you get a noise influence, then can you identify how it is coupling in and how you have managed it so far. Always good to know where your starting from in order to estimate the path to take.

    Ciao, Tim
     
  5. ue418

    Thread Starter New Member

    Jun 22, 2010
    17
    1
    Nanovate: I can't do what you suggest, as by definition the excitation signals and the return signals from the loadcell must take this path. Well, actually...now that I think about it, I could move the inamp to that far side of my envisioned long circuit board. That way, I minimize the distance the low-voltage signals must travel to-and-from the cell. Thanks.

    Tim: It's doing pretty well currently, except there is a very strange "constant offset voltage", usually negative, that occurs as soon as the windings are turned ON, but before the stepping begins. So I figure that this has to be from an electrostatic field, as only a varying magnetic field can induce a voltage in a wire/circuit. But the fluctuating waves that would come from all the stepping (steps taken every ms or so) are not seen. Of course, this could simply be due to my having an 8th-order LowPass filter downstream, before the DAQ sees the signal! I have a very low cutoff frequency of about 25 Hz. I have two separate grounds, one for my low-voltage signals and one that I call my "motor ground". This one is simply connected to the (-) terminal of my 24 V battery that drives the motor/driver.

    I have learned that this "constant offset voltage" dissipates almost instantaneously as soon as the motor begins stepping. This is very fortunate of course. If it was still dissipating - or in the process of dissipating - then there would be the problem of adding/subtracting this dying voltage from all my force/voltage results. And also, I finally found that this very fast dissipation appeared to be greatly improved once I connected my "motor ground" directly to the metal housing of the motor! In my mind, this allows that inevitable built-up "offset voltage" to quickly dissipate back to the motor ground once stepping begins.
     
  6. Ghar

    Active Member

    Mar 8, 2010
    655
    72
    If you get a constant offset you could have some common impedance coupling, where some of the motor current is going through some conductors involved in the signal circuitry.
    I'm not sure why it would go away in that scenario though... it would depend on how exactly it's connected. When you begin switching it could create an alternative path reducing the common portion. If you have a schematic it would prove or disprove this.

    I'm not sure that an electrostatic field can actually do this... an electrostatic field is conservative and the induced voltage cannot drive continuous current which is required for a DC voltage across a conducting element. An electrostatic field would make part of the outer surface of a wire positive with respect to another part of the outer surface but should not be along the wire so it shouldn't affect the circuit.
    Putting in a piece of aluminum foil (could try it connected to the circuit or not) between the motor and the signal circuit should vary this offset if it was electrostatic.
    I could be wrong...

    How big is this offset?

    Another option for a constant offset is time varying interference that is being rectified by something in the circuit.
     
  7. ue418

    Thread Starter New Member

    Jun 22, 2010
    17
    1
    Well, I do have a 6 foot cable, which is actually two "subcables", leading to the machine from another place. Each subcable is foil-shielded, with one carrying the sensitive signals (including the excitation voltages for the loadcell, and the output voltage coming out of the inamp) and the other carrying the 6 motor leads. The shielding for the latter subcable is of course connected to that "motor ground" I mentioned before. The shielding for the "loadcell signals" subcable is connected to that "signal ground". So perhaps, since these two cables run alongside each other for six feet, this is where some "common impedance coupling" could occur?? Can this phenomenon occur through air/space, or does there have to be metal-to-metal contact somewhere? Pardon my ignorance about all that.

    But regardless, I would like to know if I should use better cables in my future machines. I mean, I think I have read that there are better designs to reduce EMI than foil-shielded cables. I want these new cables to look pretty nice too. Is there something called "braided" that would perform better? If anyone has advice on better cables - with regards to protecting against EMI - I would love to hear it. Obviously the "motor" subcable should minimize outgoing EMI, while the "signal" cable should keep electromagnetic radiation (including that emanating from the adjacent motor cable) out. Again, I know my language is primitive with regards to many of these topics, including EMI issues. So I hope I am making myself clear enough. Thanks to everyone responding so far.
     
  8. Ghar

    Active Member

    Mar 8, 2010
    655
    72
    Common impedance is physically sharing conductors so the motor current would need to pass through the same conductors as the signal current in some portion of your wiring.
    Cables being beside each other can only induce time varying components without DC because of what you said (unless electrostatic works but I doubt it... haven't really come across any mention of it either). The other option like I said is a time varying component being rectified which can happen from any random transistor, diode or IC (many ICs have ESD protection diodes for example).

    What kind of frequencies are you dealing with here? Is this motor just getting DC until you start switching?
    If it is only DC at this point then I think you have to have common impedance coupling. Trace the current path through the motor and through your signals and see if they share any wire.

    When you do start switching do you apply positive and negative (bipolar) or just positive, and with what duty cycle? Low duty cycles with one polarity or about 50% with positive and negative would significantly reduce the DC offsets which could explain why it disappears when you begin to switch.

    If the motor housing is electrically connected to the windings and you connected it back to your ground then you added another current path which would take some of the motor current reducing the current that could couple to the signal circuitry. Can you confirm whether such a connection exists?

    In a system I'm working with right now we're running many things off of a battery with about 10 inches of wire before we split it apart to our devices. This wire's impedance allows a nasty 30 MHz noise from the input of a DC-DC converter to transfer to everything else in the system since it's imposed on the battery. A filter between the battery and the DC-DC stops it though getting rid of that long inductive wire should help significantly.


    As for shielding, since this is a motor we're talking about you will have relatively high current which usually implies magnetic field coupling. A non-magnetic shield doesn't really help against magnetic field coupling unless the interference is of a sufficiently high frequency (generally over a few kilohertz or so).

    Shields work the same way for emitting and receiving interference.

    There is something called a braid, it's like a mesh (braid) of copper wrapped into a tube. They come with different percent coverage. Higher percentage works to higher frequencies.
    Braid alone works at lower frequencies (both lower minimum and maximum useful frequency) than just foil. Braid is also stronger mechanically.
    Braid with foil is the best kind of shielding since you get the combination of both but the foil only really contributes for high frequency (approaching megahertz).

    Is there some sort of enclosure or chassis here? Shields are ideally an extension of your enclosure (i.e. your cables effectively never leave the box) so the ideal termination is actually the enclosure.
    An important point is that you need a 360 degree shield termination for maximum effectiveness. If that's impossible you need to keep the shield's drain wire as short as possible.

    Of course I'm assuming there isn't some bug in your DAQ or system that could cause an erroneous reading.
     
    Last edited: Jun 23, 2010
  9. timrobbins

    Active Member

    Aug 29, 2009
    318
    16
    Have you done some replacement tests such as replace the cell with a floating resistor, or replace the far end circuit with a fully shielded local end circuit - to possibly indicate where or how the offset signal is being coupled in?

    The offset may be inherent in you IA to ADC local circuit.

    Ciao, Tim
     
  10. ue418

    Thread Starter New Member

    Jun 22, 2010
    17
    1
    I know that there are no shared wires between my "loadcell" circuitry and the motor circuitry. I used to have a common ground between the two, but finally realized that this led to devastating results in terms of horrible voltages induced in my DAQ DURING stepping of the motor. Having two separate grounds is essential.

    I turn the motor windings ON for some minimal amount of time (100 ms or so), before stepping begins, per manufacturer's request. This puts the motor in a "hold current" situation, where it can resist a significant torque without turning. I would classify this as purely DC...until the stepping begins.

    I have a Unipolar, two-phase motor (six leads), with an associated driver. The control signals are all created by me on the computer...sent out via parallel port. I am using Half-stepping as well. Again, it seems everything is fine once the stepping begins. But as soon as the windings are turned ON (to the Holding Torque situation just prior to stepping), this negative voltage appears. It can't be much voltage...approximately 0.25 mV at the loadcell output (i.e., prior to the amplification). But since I amplify by about 400, then multiply by a "grams per volt" factor of around 200, it shows up as about 20 grams-force. But again, by all accounts, it fortunately "goes away" as soon as stepping begins! But it still bothers me.

    As to wheter or not the motor housing is electrically connected to the windings, I am not positive. But I did take voltage readings on the housing when in this "Hold Current" state a few months back, and couldn't read anything. So maybe this indicates that the housing is not electrically connected to the windings?

    I do have an aluminum enclosure around my motor driver, which is on the other end of the 6 foot cables. And I've connected this to my "motor ground" (the (-) terminal of my 24V motor battery). My "signal" ground is simply the (-) terminal of a 12 V battery that I use to power all of my "signal" stuff (filter, instrumentation amp, etc). I have a couple of SIP voltage converters that receive this raw 12V and create +/- 12 and +5, all regulated.
     
  11. retched

    AAC Fanatic!

    Dec 5, 2009
    5,201
    312
    You can do a continuity test with the black lead on the casing and touch the red lead to the winding leads. If you get a closed circuit between any of those, except a ground, it is usually a bad sign.
     
  12. Ghar

    Active Member

    Mar 8, 2010
    655
    72
    How did you implement this separate ground?
    What kind of connections exist between these two battery systems?

    If the grounds are connected at more than a single point then the wires are in parallel for all currents and you will have coupling.
    You can test this with a continuity check as well - disconnect it where you think you have the single point connection (at the power source?) and see if it still beeps when touching the two grounds with your leads.

    The reason I'm so insistent on this is because I can't think of what else can actually cause DC interference.
    It's very easy to accidentally connect things through mounting screws or connectors or whatever.
    A few months ago I had a fault path through a USB cable's shield connecting my circuit to the chassis where I didn't expect it.

    Have you measured this at the amplifier or are you inferring from the force reading?
     
    Last edited: Jun 24, 2010
  13. ue418

    Thread Starter New Member

    Jun 22, 2010
    17
    1
    On my 24 V "motor" battery, I simply declare its (-) terminal as "motor ground". This goes only to the GND (-) connection on the driver (right next to the (+) 24V connection). It also is connected to the cable shielding along the 6-foot "motor leads" sub-cable that I mentioned before. Those are the only two things that connect to this "motor ground". And the 24V terminal is only connected to the 24V (+) power terminal of the driver. This is the power that ultimately goes to the windings of the motor.

    The "signal ground" connects to many more things: various components and various shielding. And the same for the 12V "signal" power. Well, it first goes to the two DC/DC converters, and the regulated voltages go to the various components. I have taken great pains to separate the two grounds. And if the two positives (i.e. the 24 and 12) were somehow connected, I think many terrible things would be happening.

    I would like to better understand your statement above: "If the grounds are connected at more than a single point then the wires are in parallel for all currents and you will have coupling". What exactly were you referring to when you said "the grounds"? Did you mean, if my "motor ground" is connected to my "signal ground"? Or did you mean that if you look simply at, say, the signal ground, it should only be connected to one thing? I'm just not quite sure what you meant. I hope you can clarify for me.

    Oh, since we seem to be focusing on possible interaction between my "motor" and my "signal" circuitry, I should say that the four control inputs of the driver are fed by my "signal" 12V. But these are photocouplers and are supposedly isolated from the power (24V) portion of the driver. Just to explain these, take the Pulse input for example. My regulated 12 V goes to the (+) Pulse, then the (-) Pulse input goes through a small resistor, and then on to "signal" ground, but only after going to the Collector of my controlling transistor. So when I send a voltage to the Base of the transistor, current flows from collector to emitter (ground), which the photocoupler detects and thus moves the motor one step. So again, it seems that if my 12V was somehow interacting with the 24V within the driver, then many other bad things would be happening when the motor steps? Ironically, I used to control the driver with the same 24V, but that prevented me from doing many things down the line, and also caused my laptop parallel port ground to be at the "motor ground". All that caused design problems for me, in various ways. So a couple years ago, I started controlling the pulse, etc. with my 12 V "signal" voltage. Thanks, Ghar.
     
  14. ue418

    Thread Starter New Member

    Jun 22, 2010
    17
    1
    I don't know if I have done that or not, but it sounds like a pretty easy thing. And it sounds like a pretty clear-cut yes or no thing. I will try to do this one of these days, just to make sure it's okay. Thanks!
     
  15. Ghar

    Active Member

    Mar 8, 2010
    655
    72
    I agree, the positives can't be connected if it's working at all.
    The various shielding is the suspect to me. The computer almost definitely connects the computer chassis to the parallel port ground. Most computer cables are at least foiled and connected to the connector shells.
    If you take your computer's parallel cable to the motor driver and connect to your enclosure, you are connecting your signal ground to the motor driver chassis.
    I know you said the motor cable shield is only connected at the battery side but are you certain of that?

    If the signal subcable and motor subcable run together can their shields ever touch? That would also connect your grounds.

    I think this is the sort of thing you need to actually probe with the continuity tester. While touching your two grounds you will hear a beep. Keep disconnecting things until the beep stops. Better yet look at the resistance value, you could have a connection but with a decent resistance.

    The connection doesn't necessarily have to be the ground but it's the most likely. You could probe for resistance between various parts of your signal and motor driver circuits.

    When I say connected I refer to the two grounds being connected together at more than one point.
    This would mean the signal ground conductors and motor ground conductors are really the same current path because they're in parallel. This allows coupling between all your circuits.
     
Loading...