Ground loops

[zappedagain]

I have a circuit where three amplifiers end up in a ground loop with the power supplies. An interesting thing is that when I pull 5A through my power leads and get a 100mV ground 'bounce', my amplifiers outputs shift by 200mV. I expect the PSRR would have kept this at the same 'bounce' voltage.

Note that I'm using 'bounce' loosely here, as this is a DC effect.

The three amps are independent driving three separate loads. One has a -3X gain, and the other are 1X and they all shift by the same amount (+/-20%).

Can you suggest any other parameters I should check?

John D

 

[ScottyUK]

Post a circuit diagram and show the power routing as it actually is built. We haven't much to work with at present.


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[zappedagain]

The first page shows the power wiring. My 14 inches of #18 power wire now has 14.55" of #28 and 13" of #22 (2 conductors) in parallel with it. My signals are routed along the 14.55" of #28 so that is why I see the offset when 5A of current flows through the power wire and the new shunt.

The second page shows the amplifiers. I simplified it to only the output stages as the effect is the same with these inputs grounded.

John D

 

[VEBill]

First thing I'd try is a single point ground right in the middle.

If that didn't improve things, then I'd put on the thinking cap.

 

[djs]

Look at using differential output signals. Issue is that single ended outputs use ground as reference. If gound varies, so does the output signal.

 

[geekEE]

When you say there are shifts of 100mV or 200mV, what is your ground reference point for that measurement? Similarly, does that power supply have the grounds tied to the earth ground? Are you measuring using a scope with a standard probe or a diff probe or are you using a meter?

Glenn

 

[ScottyUK]

U12 - do you really have a 4k7 resistor in the +5V supply?

Don't equate the inherent CMRR of the IC as being that of the circuit. Component tolerances quickly degrade the CMRR of the overall circuit unless you're really careful.

VE1BLL has identified the easiest starting point - a star ground will at least eliminate interaction between circuit blocks and keep high currents out of the signal ground path. Do that and then see what happens.


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[DanEE]

VE1BLL has identified the easiest starting point - a star ground will at least eliminate interaction between circuit blocks and keep high currents out of the signal ground path. Do that and then see what happens.

Ditto.... Even on a large system having long cables, the above concept was the best and simplest of several fix attempts tried.

In this case it was AWG #6 green wire bonding that ran parallel all unbalanced audio cables.. Took out all discernable ground loop noise.

 

[zappedagain]

ScottyUK, that's 4.7 ohms (0.0047K) at U12. Good point on the CMRR; the passives can degrade it rapidly. But -6 dB instead of +60 dB???

Thanks for the ideas all. In hindsight differential signals would have been great for this circuit.

This ground loop has worked acceptably for multiple systems. Now I have a single system that is showing the problem. Interestingly enough it acted up last May too. It seems environmental effects but this system was on loan to a customer in Japan this May; last May it was in North America.

As a quick band-aid I'll try shifting the #22 ground wire to the other end of the #18 so my A/D and the sensor electronics will all ride together on the ground noise. That still doesn't get to the root cause but might let me ingore it for a while longer :S

John D

 

[VEBill]

The other thing worth keeping in mind about loops is that they can make good antennas (i.e. area inside the loop, EMI/EMC problems). Especially worth thinking about if you have a system that works 'here', but not 'there'.

In general, we would never design a system with ground loops. We would explicitly make sure that the subsystems and systems have single point grounds. And we would not allow any power supply return current to share a conductor with any signal return. Not to say mistakes aren't made, but these issues are covered by explicit design rules.

Good luck. Road trip to Japan...? ;-)

 

[djs]

Different countries have different grounding rules. For example in Europe, the norm is to bond or tie together all grounds in a complex. In the US the norm is to have each sytem independly grounded. They are not bonded together. This makes for some interesting issues when using European equipment and designs in the US.

The solution for most industrial applications is to use fiber optical cable.

 

[zappedagain]

Well my rewiring fix describe above did the trick; it dropped the ground loop caused voltage offset from 300mV to 60mV. That's tolerable for now until the system can get redesigned.

Thanks to everybody for your help!

John D