Advertisement

Blog

Ground Is a Cruel Joke

Understanding grounding structures are important to enable your circuits to work well and sometimes even to work at all. Proper grounding of circuits is very important to achieving the desired results. As a young engineer, I remember a saying by one of the senior engineers that has stuck with me my whole career. He said, “Ground is a cruel joke.” This one phrase captures the intent of this blog. I would like to discuss grounding and its effects on circuits. Ground is not always ground. Ground can be at zero volts, or it may be an analog ground where the ground potential is volts above earth ground. Such rises in potential can be intentional or unwanted.

These changes in ground are often the cause of reduced performance and sometimes can cause your circuits to not work. So the question is how we avoid grounding issues. I will provide some of my thoughts on the subject, but realize that, just as with my last blog, Capacitors: Bigger Is Not Always Better, to effectively make your grounding system work, you must understand where sensitivities occur in your circuits. With that being said, let's discuss some possibilities to avoid grounding issues.

The first thing to remember is that ground is a relative term, and under specific conditions, ground may not be what you anticipated. Under specific unwanted conditions during operation of your circuit or another circuit on your board, your circuit ground may not provide the needed current path. This condition can be the result of ground rise from currents combined into a single point ground, or it can result from the fact that the grounds used for your circuits are frequency dependent. Therefore, it is best to think of ground as a complex impedance and model it as such when assessing ground effects in your circuits.

As I said, grounding is frequency dependent, and it is important to remember this when setting up grounds for return currents. In addition, it is important to understand the frequencies presented in the circuit you are developing and how the ground for this circuit will react with the other grounds on the board and/or integrated circuit. Furthermore, it is important to understand the various ground paths that exist inside your system. The system can be an integrated IC or a development on a printed circuit board. Ground paths determine where currents flow and influence the interference you might see in your circuits or how your circuits may influence other circuits in your system.

Remember that the current must return to its source, and the current will take the lowest impedance ground path you provide, so provide one that does not influence circuits in its path. This means that the return current ground paths you provide must be over the frequency bandwidth where your circuit shows sensitivity, as well as over the bandwidth of interest to your system performance criteria.

Shown in Figures 1 and 2 are two commonly implemented grounding systems. Figure 1 shows a ground where a single point is the common ground point for many circuit blocks. In this scenario, multiple currents share a single common ground and can result in unwanted ground rise that can affect your circuit blocks. As I said, ground is frequency dependent. Therefore, a signal with high frequency transients can not only cause your circuits that share the ground to experience these transients but possibly result in unwanted mixing of these frequency components into your circuit's bandwidth of interest. Therefore, in cases where you are aware of sensitive circuits to mixing, it would be best to implement your ground system as shown in Figure 2, where each ground is separated and connected to a common ground plane.

Figure 1

Common ground point shared for multiple circuits.

Common ground point shared for multiple circuits.

Figure 2

Separate grounds to ground plane.

Separate grounds to ground plane.

As shown in Figure 2, the best ground is a ground plane, provided the plane is truly the common ground for your system. The use of such a plane provides an infinite source of recombination for holes and electrons. When you do not provide such a situation, currents can travel in unwanted directions. With these simple ideas in mind, I have outlined a small checklist to keep in mind when setting up your grounding system:

  1. Keep ground and signal leads as short as possible.
  2. Clock signal loop areas should be kept as close to zero as possible.
  3. Return current path must be kept local and short. Follow the ideas outlined in my last blog on capacitors, and use local decoupling.
  4. Long static lines can pick up noise and need to be kept as short as possible.
  5. Do not share grounds that present high frequency transients with other sensitive circuit grounds.
  6. If possible, use a package that provides low inductance to ground and be aware of the pins that present the most inductance.
  7. For choice of package pins, choose pins for grounds that provide the best and lowest impedance for the circuits that are sensitive.
  8. Understand frequencies in your system, and avoid ground loops that can be a source of coupling into your ground or transmission of your signals to other circuits.

One last item: Whether it is designing a board for implementing circuits or floor planning an integrated circuit, the path of current is important. The idea of separate ground planes in board design does not always provide the best outcome. Sometimes it won't work and will seriously degrade performance. As I stated in my last blog, if you can keep the circulating currents local to the block, this is the best situation for grounding, and here separate grounds may be more appropriate.

Do you have other items that I have missed for providing appropriate grounds in your circuits?

Have you had any experiences where the grounding system put in place caused unwanted distortion or interference into your circuits?

17 comments on “Ground Is a Cruel Joke

  1. eafpres
    September 20, 2014

    @Brandt–your point ” the fact that the grounds used for your circuits are frequency dependent” is paramount in RF systems, generally taken to be over 1 GHz.

    The problem now is that we have 25Gbps systems, say in base station backplanes, that are supposed to be square waves, and are more or less ideal depending on the design and quality of the system.  Another blog here taught me that in addition to worrying over closing the eye, if you consider the square wave comprised of an infinite series of sine waves, and signifcant power can exist in the higher harmonic frequencies.

    Under these conditions, almost anything can be an antenna, and ground planes can radiate.

  2. samicksha
    September 21, 2014

    I guess capacitance to ground is an important factor in effectiveness of signal grounds.

  3. Davidled
    September 21, 2014

    There are a few ground cases when designing circuit: chassis ground, board ground, high current ground and negative symbol.  In a few cases, negative symbol represents the communication signal, not ground.

      

  4. David Maciel Silva
    September 22, 2014

    Often the ground plane inverted mainly applied to circuits in telecommunications, where the mesh is made in the feed plane.

    Reminds me of an Integrated, a DC-DC converter LM2575HV that for maximum efficiency in the circuit ground plane should be well defined.

     

     

  5. Netcrawl
    September 23, 2014

    @Daej, I agree with you, some types of transmitting antenna systems must have a good ground to operate efficiently, a good example is vertical monopole antenna, which requires a ground plane that often consists of an interconnected network of wires running radially away from the base of antenna for a distance equal to the height of antenna. 

  6. vasanjk
    September 24, 2014

    “Do not share grounds that present high frequency transients with other sensitive circuit grounds.”

    Brandt,

    Normally we have a practice of isolating some of the high frequency transient grounds like USB port grounds via suitable Ferrite beads. How effective is it? Will it be enough or we need to follow these layout guidelines strictly?

  7. amrutah
    September 24, 2014

    @DaeJ:  I agree.

    Yes we may have isolated grounds so as to reduce the interferance from one another.  For most of the Analog circuits a Reference can be a virtual ground and helpful for signal processing, but the important ground is where the current return path is.

       We may have a negative supplies, but the ground may be different (and not necessarily the most lowest voltage).

  8. amrutah
    September 24, 2014

    @Vasanjk:

       I think, the ferrite beads might pose a useful proponent when it comes to delivering a peak current requirement, but having a inductance on the ground plane will not affect your voltage behaviour (I mean't no ground bounce scenarios??).

  9. vasanjk
    September 24, 2014

    amrutah

     

    My question is specific to the ground differences created by high speed signals such as the USB. When it comes to plane layer often, we may not have the luxury of having separate routing for such grounds till the origin. I wanted to know if ferrite beads on the usb ground could help avoid such independent,return path.

  10. amrutah
    September 24, 2014

    Brandt: Thanks for the insight on ground planes.

    “Understand frequencies in your system, and avoid ground loops that can be a source of coupling into your ground or transmission of your signals to other circuits.”

       Is there a systematic way to analyse and therefore comment that a ground loop is present and a way to avoid it.

  11. amrutah
    September 24, 2014

    In one of the earlier designs, we had the ground plane of the 50MHz oscillator shared with the 20MHz IO Buffers.  There was susbtantial ground resistance present from the oscillator to the ground PAD on chip which led a ground bounce of upto 300mV on the IO buffers severely hitting the input specifications of the IO buffers.  For the next layout release, we had to re-floorplan the ground.

  12. chirshadblog
    September 25, 2014

    @amrutah: Well what about the costing ? Has it increased majorly ? 

  13. amrutah
    September 25, 2014

    @chrishadblog: Offcourse, there was a hit on cost as well as TTM of the product.

  14. dassa.an
    September 26, 2014

    @amrulah: What kind of a hit was it ? Does it have a major impact ? 

  15. geek
    September 26, 2014

    @Amrutah: What exactly do you mean when you say “virtual ground”? Is it a path in the circuit that connects to the ground but does not act as a return path and it's just there without serving any purpose?

  16. amrutah
    September 28, 2014

    “What kind of a hit was it ? Does it have a major impact ? “

       You loose time in understanding, fixing the bug and re-release.  More-over you have to convince the customer what the problem was, what measures are taken to fix and pontentially have created sense of doubt in the product (even though there are very very few to none).

       TTM, extra resource costs (man months, tester and other software licenses) get added reducing the profit of operability.

     

  17. green_is_now
    October 23, 2014

    A star ground is best except the fact that all currents must flow through it.

    This is still the case with ground planes also as one point for return off PCb is still a star ground.

    But the currents can spread and find the least path of impedance.

    if another ground current raises the local ground reference level the currents of othe returns will not addd to this path but spread out next to it on the way to star ground.

    Once near ground point, instantanous current and residual magnetic fields from the near past influence the ground unwanted voltage at the trace or plane to via or thru-hole point.

    Ground planes add resonant risk but lower total unwanted ground voltage offsets, except if resonance is severe.

    Other planes used for power delivery at VXX can make this issue of resonance much worse.

    I have found that power planes are to be avoided, unless very good reasons are provided to swamp the risks.

    A ground resonance can couple to VCC plane and visa-versa.

    This may add a layer if a ground plane is needed in lue of power plane.

    Or a hybrid where power plane is removed above oscillators and replace with a ground patch.

    Cuts in planes to shield areas from oscillattions/waves on PCB palne.

    Best to not make these // to board edges, but messes with company standards that want all part nee and in rows and columns, creating orthoganal and // reflection zones.

    Add some angle to slot and spread out energy reducing dB of oscillation.

Leave a Reply

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