Technical Article

The Why and How of Differential Signaling

November 16, 2016 by Carsten Pinkle

Learn about the important characteristics, benefits, and applications of differential signaling, as well as the proper layout techniques for differential signals.

Learn about the important characteristics, benefits, and applications of differential signaling, as well as the proper layout techniques for differential signals.

The Basics: Single-Ended and Differential Signaling

First, we have to learn some basics about what single-ended signaling is before we can go over differential signaling and its characteristics.

Single-Ended Signaling

Single-ended signaling is a simple and common way of transmitting an electrical signal from a sender to a receiver. The electrical signal is transmitted by a voltage (often a varying voltage), which is referenced to a fixed potential, usually a 0 V node referred to as "ground."

One conductor carries the signal and one conductor carries the common reference potential. The current associated with the signal travels from sender to receiver and returns to the power supply through the ground connection. If multiple signals are transmitted, the circuit will require one conductor for each signal plus one shared ground connection; thus, for example, 16 signals can be transmitted using 17 conductors.

 

Single-ended topology

Differential Signaling

Differential signaling, which is less common than single-ended signaling, employs two complementary voltage signals in order to transmit one information signal. So one information signal requires a pair of conductors; one carries the signal and the other carries the inverted signal.

 

Single-ended vs. differential: Generic timing diagram

 

The receiver extracts information by detecting the potential difference between the inverted and non-inverted signals. The two voltage signals are "balanced," meaning that they have equal amplitude and opposite polarity relative to a common-mode voltage. The return currents associated with these voltages are also balanced and thus cancel each other out; for this reason, we can say that differential signals have (ideally) zero current flowing through the ground connection.

With differential signaling, the sender and receiver don't necessarily share a common ground reference. However, the use of differential signaling does not mean that differences in ground potential between sender and receiver have no effect on the operation of the circuit.

If multiple signals are transmitted, two conductors are needed for every signal, and it is often necessary or at least beneficial to include a ground connection, even when all the signals are differential. Thus, for example, transmitting 16 signals would require 33 conductors (compared to 17 for single-ended transmission). This demonstrates an obvious disadvantage of differential signaling.

 

Differential signaling topology

Benefits of Differential Signaling

However, there are important benefits of differential signaling that can more than compensate for the increased conductor count.

No Return Current

Since we have (ideally) no return current, the ground reference becomes less important. The ground potential can even be different at the sender and receiver or moving around within a certain acceptable range. However, you need to be careful because DC-coupled differential signaling (such as USB, RS-485, CAN) generally requires a shared ground potential to ensure that the signals stay within the interface's maximum and minimum allowable common-mode voltage.

Resistance to Incoming EMI and Crosstalk

If EMI (electromagnetic interference) or crosstalk (i.e., EMI generated by nearby signals) is introduced from outside the differential conductors, it is added equally to the inverted and non-inverted signal. The receiver responds to the difference in voltage between the two signals and not to the single-ended (i.e., ground-referenced) voltage, and thus the receiver circuitry will greatly reduce the amplitude of the interference or crosstalk.

This is why differential signals are less sensitive to EMI, crosstalk, or any other noise that couples into both signals of the differential pair.

Reduction of Outgoing EMI and Crosstalk

Rapid transitions, such as the rising and falling edges of digital signals, can generate significant amounts of EMI. Both single-ended and differential signals generate EMI, but the two signals in a differential pair will create electromagnetic fields that are (ideally) equal in magnitude but opposite in polarity. This, in conjunction with techniques that maintain close proximity between the two conductors (such as the use of twisted-pair cable), ensures that the emissions from the two conductors will largely cancel each other out.

Lower-Voltage Operation

Single-ended signals must maintain a relatively high voltage to ensure adequate signal-to-noise ratio (SNR). Common single-ended interface voltages are 3.3 V and 5 V. Because of their improved resistance to noise, differential signals can use lower voltages and still maintain adequate SNR. Also, the SNR of differential signaling is automatically increased by a factor of two relative to an equivalent single-ended implementation, because the dynamic range at the differential receiver is twice as high as the dynamic range of each signal within the differential pair.

The ability to successfully transfer data using lower signal voltages comes with a few important benefits:

  • Lower supply voltages can be used.
  • Smaller voltage transitions
    • reduce radiated EMI,
    • reduce power consumption, and
    • allow for higher operating frequencies.

High or Low State and Precise Timing

Have you ever wondered how exactly we decide if a signal is in a logic-high or logic-low state? In single-ended systems, we have to consider the power supply voltage, the threshold characteristics of the receiver circuitry, perhaps the value of a reference voltage. And of course there are variations and tolerances, which bring additional uncertainty into the logic-high-or-logic-low question.

In differential signals, determining the logic state is more straightforward. If the non-inverted signal's voltage is higher than the inverted signal's voltage, you have logic high. If the non-inverted voltage is lower than the inverted voltage, you have logic low. And the transition between the two states is the point at which the non-inverted and inverted signals intersect—i.e., the crossover point.

This is one reason why it is important to match the lengths of wires or traces carrying differential signals: For maximum timing precision, you want the crossover point to correspond exactly to the logic transition, but when the two conductors in the pair are not of equal length, the difference in propagation delay will cause the crossover point to shift.

Applications

There are currently many interface standards that employ differential signals. These include the following:

Clearly, the theoretical advantages of differential signaling have been confirmed by practical use in countless real-world applications.

Basic PCB Techniques for Routing Differential Traces

Finally, let's learn the basics of how differential traces are routed on PCBs. Routing differential signals can be a bit complex, but there are some basic rules that make the process more straightforward.

Length and Length Matching – Keep It Equal!

Differential signals are (ideally) equal in magnitude and opposite in polarity. Thus, in the ideal case, no net return current will flow through ground. This absence of return current is a good thing, so we want to keep everything as ideal as possible, and that means we need equal lengths for the two traces in a differential pair.

The higher the rise/fall time of your signal (not to be confused with the frequency of the signal), the more you have to ensure that the traces have identical length. Your layout program might include a feature that helps you to fine-tune the length of traces for differential pairs. If you're having difficulty achieving equal length, you can use the "meander" technique.

 

An example of a meandered trace

Width and Spacing – Keep It Constant!

The closer the differential conductors are, the better the coupling of the signals will be. Generated EMI will cancel out more effectively, and received EMI will couple more equally into both signals. So try to bring them really close together.

You should route the differential-pair conductors as far away from neighboring signals as possible, in order to avoid interference. The width of and space between your traces should be selected according to the target impedance and should stay constant over the entire length of the traces. So if possible, the traces should remain parallel as they travel around the PCB.

Impedance – Minimize Variations!

One of the most important things to do when designing a PCB with differential signals is to find out the target impedance for your application and then lay out your differential pairs accordingly. Also, keep impedance variations as small as possible.

The impedance of your differential line depends on factors such as the width of the trace, the coupling of the traces, the copper's thickness, and the PCB's material and layer stack-up. Consider each of these as you try to avoid anything that changes the impedance of your differential pair.

Do not route high-speed signals over a gap between copper areas on a plane layer, because this also affects your impedance. Try to avoid discontinuities in ground planes.

Layout Recommendations – Read, Analyze, and Overthink Them!

And, last but not least, there is one very important thing you have to do when routing differential traces: Get the datasheet and/or application notes for the chip that is sending or receiving the differential signal, read through the layout recommendations, and analyze them closely. This way you can implement the best possible layout within the constraints of a particular design.

Conclusion

Differential signaling allows us to transmit information with lower voltages, good SNR, improved immunity to noise, and higher data rates. On the other hand, the conductor count increases, and the system will need specialized transmitters and receivers instead of standard digital ICs.

Nowadays, differential signals are part of many standards, including LVDS, USB, CAN, RS-485, and Ethernet, and thus we all should be (at the very least) familiar with this technology. If you are actually designing a PCB with differential signals, remember to consult relevant datasheets and app notes, and if necessary read this article again!

 

Featured image courtesy of Hardwareonkel (own work) [CC BY-SA 3.0]
5 Comments
  • C
    cutedaddy November 28, 2016

    Hi Carsten, tx for a great article, curt, clear and instructive.
    Hope this is not a dumb question, but i wonder why you mention “SNR automatically increased by a factor of two” kind of as an afterthought, while it would seem an essential advantage of balanced vs. SE in, for instance, many audio applications. Regards, Yuri

    Gonna read more of you. Can i quote you when musing on audio (w/ref. of course)?

    Like. Reply
  • C
    carstenp November 29, 2016

    Hi Yuri,

    Thanks for your feedback!

    I think dumb questions doesn’t exist, only people who don’t dare to ask. So feel free to ask!

    Yes you’re right, the improved SNR is an essential advantage of differential signals, especially for analog audio applications.

    But in this section I want to point out the digital part of DS. In order to get higher data rates, less power consumption, less charging capacitance, etc., you reduce the voltage of the differential signal. If you lower the voltage, your snr gets smaller.
    In analog audio application, it’s a different approach. You want to have a high snr and you don’t intentionally want to reduce it, because you loose headroom. I don’t know if that answers your question?

    I don’t know if I am a good reference to quote. I’m just a little light.. But if you want to, feel free to do it.

    Cheers Carsten

     

    Like. Reply
  • J
    johnhardinece June 04, 2017

    A comment and a question:

    Comment:  Differential signaling is by no means less common than single ended signaling.  In fact, it is very common, especially when transmitting a digital signal from one board to another and from chip to chip to chip on the same board for high speed data transfer.  Take a look at how HSS (High Speed SERDES) are implemented, as well as many oscillators use differential signaling to provide clean clocks.  While single ended signaling is still very common, and I’d imagine that will not change anytime soon, differential signaling is too.  As with everything in engineering, the method used to implement a solution depends on the problem that needs to be solved, and both single ended and differential signaling are just tools in the toolbox.

    Question: If the positive leg of a differential signal doesn’t have the same ground reference, then how can the benefits of differential signaling be accomplished with regard to noise cancelation, fields cancelation, etc?  If both legs are referencing a different ground, then one cannot assume both grounds are at 0V, thus eliminating the common mode rejection benefit as well.  LVDS drivers may indeed have more than one ground pin, but please, do not connect different grounds to them!  Also, while the theory of differential signaling is important, only those designing LVDS drivers are going to need implement them directly, everyone else is going to use them from the perspective of a user.  And in the case of a user, a single common ground is definitely what I’d recommend over using different groundss.

    Thanks,
    John

    Like. Reply