Signal Chain Basics #84: Why RS485 Does Not Need Ground Wires

RS485 is the only differential interface allowing for robust data transmission over miles — long distances at reasonably high data rates of up to 100kbps. Preferably bus nodes are connected by daisy-chain, or a single cable run that connects one node to the next. Because each node receives its power supply from a different point in the electrical installation, ground potential differences (GPDs) between bus nodes are common.

Figure 1

RS485 bus with ground potential differences between bus nodes

RS485 bus with ground potential differences between bus nodes

These GPDs increase the common-mode voltage on the bus and can cause data errors or device damage, if not properly dealt with. RS485-compliant transceivers tolerate GPDs of up to ±7V. Higher GPDs however require special transceivers with a higher input voltage range, or galvanically isolated transceivers.

Some engineers use ground wires between bus nodes in order to tie the transceiver grounds to a defined voltage potential. This method can have dangerous outcomes as it shortens multiple ground potential differences together causing prohibitively large ground currents that literally burn ground wires.

On the basis of a single point-to-point interface, this article shows that differential transceivers are designed to operate without ground wires.

The differential driver has a H-bridge output stage that drives current from terminal A to terminal B and vice versa, depending on the logic states at data input, D (Figure 2).

Figure 2

H-bridge output stage

H-bridge output stage

Because the line voltages, VA and VB , switch between two positive potentials, the driver can be modeled as a signal source with a common-mode output offset, VOS , and a superimposed differential voltage of ±VOD /2 (Figure 3).

Figure 3

Driver model and line voltages

Driver model and line voltages

The differential receiver’s input stage consists of a resistor network that attenuates large input signals and biases the comparator inputs with respect to receiver ground (Figure 4). By referencing the inputs to receiver ground, the receiver can build the voltage difference between VA and VB without a ground connection between driver and receiver ground.

Figure 4

Differential receiver input stage

Differential receiver input stage

Voltage divider action between the input resistors (Rin) and the bias resistors (Rb) attenuates the input signal (common-mode and differential voltage components likewise) by a factor of ten. This prevents the comparator inputs from saturation (Figure 5).

Figure 5

Resistor network attenuates differential and common-mode signals

Resistor network attenuates differential and common-mode signals

Voltage divider action between the bias resistors biases the comparator inputs with VCC /2 potential, thus enabling the receiver to work from a single-supply voltage.

The data link in Figure 6 applies the previously gained information by showing the GPD as a common-mode voltage that adds to the driver output offset. In order for the data link to work reliably, the sum of the total common-mode voltage and the maximum signal must not exceed the specified maximum receiver input voltage:

VIN-max ≥ VCM + VD /2     (1)

Figure 6

Point-to-point data link with common-mode and signal voltages

Point-to-point data link with common-mode and signal voltages

In an RS485 data link the driver offset typically is half the driver supply, e.g., 2.5V. Adding a GPD of ±7V yields a common-mode voltage range from 9.5V to −4.5V. A superimposed differential signal of ±1V yields the final bus voltage swing from 10.5V to −5.5V.

In comparison, EIA-485 specifies a common-mode range from 12V to −7V. Device data sheets commonly define this range as the recommended receiver input voltage range, while specifying the absolute maximum input voltage range from to 14V to −9V.

Industrial networks often experience ground potential differences significantly larger than 7V. For GPDs up to 25V, transceivers with high common-mode input ranges are recommended. GPDs higher than that require bus transceivers to be electrically isolated from their local node circuitry by the means of galvanic isolators in the signal and supply lines (Figure 7).

This method leaves all transceiver grounds floating, thus reducing the GPD down to 0V and the receiver input common-mode voltage to half the receiver supply, VCC2-ISO /2.

The correct method for designing a differential data link is without ground wires. For ground potential differences higher than the specified ±7V in EIA-485, use transceivers with high-common-mode capability, or isolated bus transceivers.

Figure 7

Isolated data links drastically reduce VCM and GPD

Isolated data links drastically reduce VCM and GPD

Please join us next time when we will discuss the difference between gain bandwidth product and unity gain bandwidth.

For more information about RS485, see RS485 Design Support . About the Author
Thomas Kugelstadt is a senior systems engineer with Texas Instruments. He is responsible for defining new, high-performance analog products and developing complete system solutions for industrial interfaces with robust transient protection. He is a Graduate Engineer from the Frankfurt University of Applied Science. Thomas can be reached at . Related posts:

23 comments on “Signal Chain Basics #84: Why RS485 Does Not Need Ground Wires

  1. Davidled
    December 5, 2013

    Firstly, there is no information for distance limit of RS485. No ground wires would get caution for PCB interface such as transceiver ground floating. It might require ground isolation among RS485, signal condition circuitry and high voltage current if it requires. I think that engineers need to get the allowable maximum distance between transceiver and receiver. Secondly, I wonder what maximum of total number of node could be used in the RS484 interface.

  2. BIG TOM_#1
    December 5, 2013

    600 words limit does not allow for deep explanations. then TSB89 (RS-485 application bulletin) gives a mx distance with 4000 ft AWG24 cable. At this length the wire resistance approaches the value of the termination resistor and attenuates the signal by half (-6dB).

    the max number of nodes depends on their unit loads. The standard defines a maximum load of 32 unit loads. If you use a 1UL transceiver you can connect 32 of them on a bus. If you use 1/8 UL transceivers, you canconnect 8 x 32 = 256 on the bus. well, that's what the theory. In paxice the number of nodes also depends on the transceiver capacitance , the node capacitance, the cable capacitance, the distance betwee node, and many other beautiful things.

    With regards to floating grounds, I showed and explained that differential interfaces not just don't need but rather must not use ground wires. Someone wants to state the opposite, document your “proof” and send it to .

    just talking very vaguely about it does not help much.

    thank you, Thomas

  3. BillWM
    December 6, 2013

    Do you know of any good app notes for XFMR coupled manchester data over RS485 — solves dc ballance issue.

  4. Vishal Prajapati
    December 7, 2013

    @William, Don't you think that instead of isolation transformer, opto isolators or even magnetic isolators are good bet for digital data transmission? Opto are cheaper, lighter and smaller than transformers.

  5. BillWM
    December 7, 2013

    Opto's have lightning induced transient protection costs that are far more than those for a coupling transformer based circuit — not to say that either way will not work, just that unless one replaces the optocoupler driving wire with a full fiber link the optocooupler i/o requires protection circuits to provide full protection like the 485 driver driving the transformer does.


    (Fiber itself can be suceptible to fluids and thus be an issue where electro-hydraulic actuators are required)

  6. Davidled
    December 7, 2013

    Opto isolator (chip) has a limited voltage range, but cost is inexpensive. Isolation transformer is better stable for high powered driven LED, even though the range of both inductor and voltage is variable for a different isolation transformer.

  7. samicksha
    December 7, 2013

    Yes its basic concept clears the purpose, i.e main function of an opto-isolator is to block such high voltages and voltage transients, so that a surge in one part of the system will not disrupt or destroy the other parts.

  8. Victor Lorenzo
    December 9, 2013

    @Thomas, thanks for the post, it covers several key aspects about RS485 and mentions several core concepts when interconnecting data networks with this type of interface.

    The only part I miss here is the inputs protection, which is imperative to be addressed by the designer.

  9. Victor Lorenzo
    December 9, 2013

    @William: “Do you know of any good app notes for XFMR coupled manchester data over RS485

    I'm just curious about that. Would you please comment a little bit about it, specially about the need for the XFMR?

  10. Victor Lorenzo
    December 9, 2013

    @DaeJ, “Opto isolator (chip) has a limited voltage range, but cost is inexpensive

    It all depends on many factors, there's a number of pulse transformers with a fairly small footprint and relatively high bandwith (although this is probably not the most appropiate term) available for less than 3$, and 'good' optocouplers with high isolation voltaje and rated at 5V/us or more for industrial applications tend to be in the range from 2$ and up.

    I think at the end its all up-to the application we're working on. The good news here is we have plenty of options to chose from.

  11. BillWM
    December 9, 2013

    The real issue without the Manchester/Transformer arrangement is the DC ballance issue where a long string on 1's or 0's causes a shift in the voltage level on the reciever inputs when long wires with resistance are used — this can cause data to be interpreted as the wrong value or level when no transformer / manchester code is used.

  12. Victor Lorenzo
    December 9, 2013

    @William, Please, feel free to correct me if I'm wrong. At both ends of the twisted pair we put termination resistors, and we tend to make both wires of the twisted pair of almost the same length. In principle, the voltage developed across both wires due to bias currents should be almost identical but opposite in polarity.

    The wires resistance contributes to reducing the effective dynamic range, but if you manage to use effectively isolated input/output stages (floating, like in the LTC1535 isolated RS485 transceived, for citing one) the effect of sending long strings of 0's or 1's should be minimum provided that you're connecting directly to the bus. If isolation is accomplished using transformers the DC ballance does show up as the transformer blocks the DC level, in which case we need data encodings like Manchester.

    Have you used another encoding besides Manchester with isolation transformers?

  13. BIG TOM_#1
    December 9, 2013

    1) sorry for my delayed response. I'm currently travelling through Asia. 2) I do not have an application note for transformer isolation with Manchester encoding. 3) transformer isolation is usually appllied on the bus side while isolation through opto-couplers is installed on the control or logic side of the transceiver, i.e. between transceiver and UART or MCU. Optos can be used at low data rates up to 100 kbps, as they are dirt cheap (4 to 5 cents). At higher data rates, such as high-speed MODBUS or Profibus applications, RS-485 transceivers with integrated digital (capacitive or magnetic) isolation are used. 

    Isolation prevents high common-mode voltages from corrupting your data traffic, i.e. by preventing the flow of ground loop currents. Isolation does NOT protect your bus node against fast transients from ESD or EFT or Surge. TI has worked with Bourns on transient protection circuits for RS-485 bus nodes. Please refer to the application section in the in the SN65HVD82 data sheet on I provide a lot of information on the individual transient energies and their suppression. Figure 23 in the data sheet has a detailed application circuit that should simplify your protection design for an isolated bus node. Regards, Thomas

  14. Victor Lorenzo
    December 9, 2013

    @Thomas, thanks for pointing us to the SN65HVD82 datasheet, as you say, it contains usefull information about transients and their suppression.

  15. BillWM
    December 9, 2013

    There are other line codings, as well as drive types — T1/E1 uses Alternate Mark Inversion to drive the transformer for example rather than Manchester (tri-level signal)  done with 485 this is either 1, Z, or 0 for drive with 1, and 0 representing a Logic High and Z representing logic 0

  16. Victor Lorenzo
    December 10, 2013

    @William, thanks for the comment about the Alternate Mark Inversion line coding. BR, Victor

  17. BIG TOM_#1
    December 11, 2013

    actually I use Manchester coding for getting hihg data rates across long cable runs. typically the higher cable capacitance cause increased loading and ISI (inter-symbol interference) can close eyes quickly. Using a clock and XOR-ing it with data gives you manchester encoded data. on the receive side you either need to use a clock recovery circuit (typically with 6 to 8 times higher frequency) for decoding, or you simply use a second differential link for getting the clock to the receiver.

    With a small RS422 IC, such as a dual driver and dual receiver IC, one can accomplish this task quite easily. On the receiver side you simply use another XOR to decode the encoded data with the received clock to get to the original data. – Thomas


  18. 204204
    March 3, 2016

    Hello gentlemen, This is Arabinda sahu. Could anybody confirm whether we can take a single RS 485 signal from the field(2 wire from the Ultrasonic meter on the field) and loop that signal into two differnet masters(Say two flow computers) in the control room.

  19. Krishnamoorthi
    August 11, 2016

    I believe 2 masters and one slave communication is not possible if both the masters are communicating at the same time.  

  20. Krishnamoorthi
    August 11, 2016

    Hi Thomas,

    Could you please explain how do you derive Driver model(Fig.3) from H-bridge(Fig.4)?

  21. BIG TOM_#1
    August 16, 2016

    This is correct, and it is called bus contention. As RS-485 is not a protocol but only an “electrical specification for drivers and receivers for balanced communication”, the higher level software must take care of traffic management to prevent two or more bus nodes from transmitting at the same time.

    Regards, Thomas ()

  22. BIG TOM_#1
    August 16, 2016

    For example, during switching one of the line voltages is: Va or Vb = Vcc – Vf – Vr-on, while the opposite line voltage is: Vb or Va = Vcc – Vf – Vr-on(Q2) – Vod

    The voltage common to both lines is therefore: Vcm = Vcc – Vf – Vr-on, while the difference voltage is Vod (or Vrd, the voltage across Rd).

    Instead of assigning Vod to one line, such as Va = Vcm and Vb = Vcm – Vod, the difference voltage is split equally to both lines, making:

    Va = Vcm + Vod/2 and Vb = Vcm – Vod/2 during one switching cycle, and

    Va = Vcm – Vod/2 and Vb = Vcm + Vod/2 during the other switching cycle.

    Hence I defined Va and Vb as:

    Va = Vcm +/- Vod/2 

    Vb = Vcm -/+ Vod/2.

    Hope this clarifies it.

    Regards, Thomas   




  23. Krishnamoorthi
    August 24, 2016

    Thanks Thomas for the clarification. Is it correct that Rd = Rt || (2Rin + Rb) ?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.