What is ground loop and how to prevent?

As you know, there are no perfect conductors in the things we build. If you are wiring a circuit, it is sometimes difficult to get all the ground wires tied to a single point, and you will occasionally end up with a loop in your ground system. In some applications, this will not present a problem. But in other cases, in particular in audio amplifiers or sensitive sensor and instrumentation amplifiers, you will have a significant difference from one end of the loop to the other. To illustrate how this causes problems, take any complex circuit board and measure the supply voltage right where the power is connected to the circuit board. Then, with the ground lead still at the ground connection, read the power supply voltage at the farthest chip on the board. You may see a half volt difference or more in some cases. Now imagine that you have a standard "star" type ground arrangement and decoupling capacitors on each chip- no problem, ground noise will be pretty much taken care of. But in a loop you end up with two problems- one is that their can be a large difference across the loop which may actually drag ground up a quarter to a half volt in parts of the circuit that otherwise would be unaffected. The other is that a loop acts as an antenna and can also pick up switching transients. For many logic applications, the only real problem you may see is either "brownout" on some chips or a loss of noise immunity- which can lead to "hair trigger" or sporadic operation. Faulty clocking or state changes can happen then. But for audio applications, the ground loop can be death- unexpected tiny noises become avalanches of audio garbage, adding pops and clicks, whines, and hum to the final output. And one final item- in some cases, a ground loop can actual resonate or oscillate. You can only imagine what this can do to sensitive components that want a clean regulated supply for operation. In summary, avoid the ground loop. It can cause vague, hard to track down problems in digital systems, and outright terrible performance in audio and instrumentation. Stick with "star" or single-point ground, and try really hard to isolate digital ground from analog ground when both are present.

Cheers!

Chip Shults My robotics, space and CGI web page -

Even with the topology that you are suggesting, there are still ground loops, they cannot be avoided until every ground is in the same point spatially There is a limit to transport speed for electrons in Cu (about 0.67c - 1ft/ns) and any other known material that we use - no system that has been built is without ground loops (the potential differences caused by currents flowing in a non-point system), we just try to minimize them and their effects. Audio is a good example, but things are happening slowly there, try getting ground loops out of a chip at 1/4 wave RF frequncies (how long is a quarter wave at 50GHz) -

0402's are bigger than that and will cause ground loops there.

Andrew

Sir Charles W. Shults III wrote:

tied

volt

much

That's what we always facing!!!, everytime we develop a board, (we deal with hi-resolution video,) we spend most of the time to fix this kind of noise. In video it causes terrible humming, windstorm, and sometime the ugly sparkles noises. I tell to the PCB guys (well I deal fpga and general R&D) always keep digital and analog ground separates, and keep in mind that the digital return currents are top killers and must be well isolated from analog section.

But you now, some how it still end up noisy, since those PCB guys tend to make an easy job. He always questions why 2,3, or more power/ground plans, why not a single ground plans for all :(, while we ussually have 3 big sections: input analog, digital (DSP/FPGA), and output analog, plus several clock/oscillator/PLL, A/D, D/A,... Anyway, a pro PCB designer must be tough/experience/and most of all he need to learn!!!

Another thing that causes problems, within analog, is supply currents for push-pull output stages. Each supply rail is supplying current primarily on only half the cycle, so its current is a very distorted version of the signal. If that current is allowed to couple back into the input - for instance, if the ground side of a supply filter capacitor shares a PCB trace with the ground lead of the input for an inch or two - then there will be substantial distortion introduced into the output.

Andrew, you are right. I did not mean to imply that this would be the end of the problem, simply that you must be very circumspect about how you lay things out. In the microwave and higher frequencies, you must use stripline design methods to even come close. Every component then has other parasitic functions, where resistors become inductors, capacitors become resistors, and inductors show capacitance. My note was aimed primarily at the "basics" type readers in the groups- those who are learning enough to make projects that work, but are plagued with odd problems that just will not go away. Now some of them may see an answer, or at least a better understanding, of what is going on.

Cheers!

Chip Shults My robotics, space and CGI web page -

I think that the problem in understanding ground loop may be that the term "ground" is so poorly understood. "Ground" means "the zero-volt reference for everything". If zero isn't really zero, you can't guarantee what will happen.

Any time you push current thru a wire, there is a voltage drop. If you have one "ground" here and another "ground" over there, there will be a potential difference (voltage). Chip's mentions that a star ground is the classic way to deal with this. I've heard this called a Mecca Ground (though a google search doesn't find this to be a common usage). Low-impedence ground planes also try to minimize this.

Andrew points to the propagation delay in circuits with fast risetimes. This is another reason for reducing distances and trying for a clean one-point ground.

Guess you had better either do your own boards, sit with the designer, or be firm with written specifications for the board.

I had not heard "sparkle" for a while - old A/D's used to do this, and I saw what you are talking about on the output stage (JFET) of Kodak CCD's (still there) - make sure that the reset pulse is doing it's job, and maybe CDS the thing. Try isolating your ouputs (cut the trace or pull the resistor, hodl some nominal voltage on the A/D and see where it's coming from. CCD's have enormous current loads on the transfer gates (some of the philips ones require more than 11A for short periods, this feeds through to the output stages if there is not enough juice available locally).

First thing I do anymore when doing anything above 30 MHz is to do the ground diagrams, max length versus current at nodes etc. Power supplies for these things have to be decoupled viciously - figure a good LC to avoid feeding it back on both the ground and the high (low) side. I found the only way to get beyond 65dB with CCD's was to generate local ground and VCC's, using good video op-amps for feedback to the power supply for isolation. The star configuration that you are talking about (multiple planes) with seperate grounds for analog and digital suggests that the digital section is not decoupled enough - I ran ECL (how do you say switch current) parts with 150 Ohm termination to ground on the same ground as an A/D (ECL at 50-3200MHz, 12 bit A/D (73db) sampling at

40MHz), no problems as long as the decoupling was done well.

Andrew

Marlboro wrote:

tied

volt

much

Example "earth loop":-

jack lead earth ------------------- | | | | |power cord earth |power cord earth | | | | mains plug mains plug | | ------------------- mains earth

60Hz radiation from e.g. a mains transformer can induce "mains hum" in the loop.

Audio signals travelling along the jack plug lead e.g. from a record player to an amplifier will now have mains hum added to them(via their earth "return" (ironically, the screened jack lead makes it worse*).

If you have an older "modular" hi-fi system with separate mains cords, you will find that only one of them has an earth, the others will only have L and N.

*because the screen *prevents* mains hum on the jack lead's central signal wire: this would otherwise *oppose* and tend to *cancel* the pickup in the screen.

Robin