Shop OBEX P1 Docs P2 Docs Learn Events
serial communication without common ground? — Parallax Forums

serial communication without common ground?

tonyp12tonyp12 Posts: 1,951
edited 2011-05-11 22:09 in General Discussion
On battery operated devise I bring two wires out for communication.
It only works if i replace the battery and get the power& gnd from same device that is sending the data.

In this case it's spy bi-wires for a TI-msp430,
but I guess this problem is in general too all 2 wire serial communication?

Comments

  • Mike GreenMike Green Posts: 23,101
    edited 2011-05-09 15:51
    Electricity has to flow in a complete circuit. Normally, the common ground serves as the second wire of the circuit while the "signal wire" acts as the first wire. If you have no common ground, you still need a second signal wire. RS485 is an example of a balanced signalling circuit used for serial communications. It uses two "signal wires" and doesn't need a common ground, although that helps with noise suppression. A "current loop" is another serial signalling convention that can work in the absence of a common ground.
  • pjvpjv Posts: 1,903
    edited 2011-05-09 16:10
    Hello Mike;

    I believe that the 485 drivers DO require a ground line in order to keep their inputs within the allowed input range of 7 ? (not certain of that value without lookup) volts. Sometimes the circuitry will work without the ground line, provided that no 485 chips are powered from isolated supplies.

    Cheers,

    Peter (pjv)
  • tonyp12tonyp12 Posts: 1,951
    edited 2011-05-09 16:37
    What would be the best way to tackle this 2wire jtag communication?
    One is clock, second is 2way I/O
    It have to be cheap/few parts, but the distance is under a 1ft

    I guess I need to start by making it a 3wire plug.
  • kwinnkwinn Posts: 8,697
    edited 2011-05-09 17:07
    tonyp12 wrote: »
    What would be the best way to tackle this 2wire jtag communication?
    One is clock, second is 2way I/O
    It have to be cheap/few parts, but the distance is under a 1ft

    I guess I need to start by making it a 3wire plug.

    3 wires would be required if the two circuits do not already share a ground since the clock and I/O are 2 distinct signals. I find "1 wire signalling" to be a bit of a misnomer. It really should be labelled " 1 wire + ground "
  • Mike GreenMike Green Posts: 23,101
    edited 2011-05-09 18:01
    Unless there's some reason to electrically isolate the two devices, the easiest connection is to have a common ground with two different power supplies, a battery for the TI-430 device and some other power source for the other device. You then have the two signal lines (clock and data) between the two devices with the common ground as the return for the signal lines.
  • Duane DegnDuane Degn Posts: 10,588
    edited 2011-05-10 12:20
    Peter,

    I agree with Mike on this one. I'm pretty sure (well really sure) that RS-485 lines do not need a common ground.

    My knowledge is pretty much all academic though. I read the book Serial Port Complete which talks a lot about RS-485 communication. I believe if you have a shielded cable on the RS-485 line that you are only supposed to ground one end of it. (I think JonnyMac also said the same thing.)

    I think not needing a common ground, is one of reasons for the popularity of RS-485.

    Duane
  • LeonLeon Posts: 7,620
    edited 2011-05-10 12:24
    It's also differential, which gives good noise immunity.
  • RobotWorkshopRobotWorkshop Posts: 2,307
    edited 2011-05-10 12:25
    Here is a pretty good site with details on connecting RS-485 and RS-232 connections:

    http://www.interfacebus.com/Design_Connector_RS485.html

    And an article on using RS-485:

    http://www.embeddedsys.com/subpages/resources/images/documents/microsys_art_RS485.pdf
  • Phil Pilgrim (PhiPi)Phil Pilgrim (PhiPi) Posts: 23,514
    edited 2011-05-10 13:49
    If two RS485/422 nodes are truly isolated from each other, without a common ground, whoever is currently transmitting on the bus establishes the common-mode level of "ground". However, if you do have a common ground wire in the network cable, each node's Vss should be connected to it via a series 100-ohm resistor in order to prevent large ground currents if there turns out to be a large common-mode potential between two nodes. For a short two-point "network", though, the resistors would not be necessary. Here's some more info:

    None of this RS485 stuff addresses the OP's concerns, though, since he has one unidirectional clock signal and one bidirectional data signal. Mike's answer is the one I'd go with.

    -Phil
  • pjvpjv Posts: 1,903
    edited 2011-05-11 09:41
    Hi All;

    Regarding a common (ground?) between RS485 devices...... the spec sheet states a common mode voltage range of +7 to -12 volts. If there are only the two differential data lines connecting two chips on a bus, then that requirement is easily exceeded. A common of some sort, although not necessarily "earth" ground, is required to keep chips within that range.

    I have first hand observation of failures for not following that spec.

    Where required, true high speed (in the megahertz range) isolation to thousands of volts can be provided by means of the ADMu chips. They use micropowered RF links accross a silicon barrier. Various options of multiple channels in both directions in a single package are available. I use these for industrial applications. They even have one that "passes through" 3.3 or 5 V DC at a few (10-ish mA), completely isolated to the same 5,000 voltage, to power "down stream" devices. I actually run some SXes that way.

    Cheers,

    Peter (pjv)
  • localrogerlocalroger Posts: 3,452
    edited 2011-05-11 15:12
    Consider switching to 20 milliamp current loop instead of RS485. Current loop is optically isolated and doesn't care about commod mode voltages because the receiver is a LED shining on a phototransistor. It's also much easier to troubleshoot by voltage readings if it's not working.
  • kwinnkwinn Posts: 8,697
    edited 2011-05-11 20:27
    localroger wrote: »
    Consider switching to 20 milliamp current loop instead of RS485. Current loop is optically isolated and doesn't care about commod mode voltages because the receiver is a LED shining on a phototransistor. It's also much easier to troubleshoot by voltage readings if it's not working.

    20 mA current loop does not have to be optically isolated but optically isolated current loop is probably the most reliable and least problematic way to go. Like other electrical signalling methods it still requires two wires, even if one of the two is a common ground. The only true one connection communication system I am aware of is optical fiber, and it can send full duplex over a single fiber. More complex and costly however.
  • max72max72 Posts: 1,155
    edited 2011-05-11 22:09
    As long as you have a common ground (all the grounds connected) you can use multiple battery packs.
    As an alternative you could investigate wireless transmission (Xbee or similar).
    Massimo
Sign In or Register to comment.