1. Home
  2. Electronics Design
  3. Help with avoiding ground-loops on my PCB+external

Help with avoiding ground-loops on my PCB+external

All,

Apologies for the cross-post, I know a lot of people who deal with FPGAs are also experts with PCBs, stack-ups, high-speed layout, hoping for some good ideas from both groups. I'm not an expert on board design or electronics, but I'm trying to do things right from the start...

My issue in a nut shell: How can I properly prevent ground-loops from external connections that interface to my board? I receive several digital signals streams from two to four external boards, single- ended, ~5 Mbps, almost zero-balanced. My board has sensitive analog and audio/video and I'd like to avoid the possibility of ground-loops,

60 Hz hum, etc.

I know of three methods to isolate and I list the methods and my thoughts.

  1. I can use a capacitor in-series with each input signal and level- shift it back so my digital ICs will properly trigger and keep the ext_GND isolated from my sys_GND. I have tested this and it's reliable for my configuration.

But! I'm very worried about EMI and return currents. If I use a capacitor to couple the external signal in AND I don't connect ext_GND(s) to my sys_GND, I fear that the return current from the input signal will see an impedance discontinuity and cause me MORE EMI headaches. I reason that the current should flow in a loop and while the capacitor isolates my two systems, the return current will still have to go back to the original source and since my GNDs are discontiguous, I have a high-impedance return (maybe through free space) and this will couple more noise into my system.

  1. I can use an isolation transformer (1:1) and a capacitor on the input side to avoid saturating the core (strip DC) and use a cap on the output & restore the DC.

I wonder if the capacitance between windings, other parasitics and EM "leakage" however won't end up defeating the purpose of isolation and will end up coupling noise from the external circuits into my system. Since my signals are single-ended I don't have a differential receiver to reject the common-mode noise either.

I'm also uncertain if the return current has a path, but based on my recollection of transformers, the load impedance is reflected back to the source, so there is a "path" and both sides are coupled together.

  1. I can use a digital isolator (NVE, ADI) and have it straddle the external input block and my digital system.

The issue here is the need for power on the external input side that would have to be isolated from the digital side. I also wonder, since the NVE/ADI parts end up using magnetic isolation techniques, will they end up broadcasting any EM?

Thank you! Sam.

Reply to
digitaljanitor
Loading thread data ...

I guess my first question is to figure out if any of this is overkill. When you say you have signals coming from "external" boards, do you mean a separate chassis? If the boards are in the same case, then proper grounding techniques should do the job.

If you are sure that you need this sort of isolation there is a third choice, optical isolation. 5 Mbps is not beyond the speed of optoisolators. A quick search turns up parts rated up to 50 Mbps. The input devices require no special circuitry and output devices simply require a power supply that is separate and isolated since this becomes the only path for conducted noise.

I guess my real concern is that if you are so worried about the return current and coupled EMI from the external digital signals, why don't you put the digital electronics on a separate card? That is the best way to provide proper isolation from the digital noise... at least the best first step.

Rick

Reply to
rickman

Please clarify exactly what type of signals you wish to isolate.

Analog Video?

Digial Video?

Analog - AC with DC component? just AC?

Other

There are many ways to isolate...for particular type of signals... so please clarify exactly what type of signals.

Yes -for digital you might try some of the AD iCoupler isolators (some even pass power 50mA to the isolated side).

Joe

Reply to
Joe G (Home)

"digitaljanitor" schreef in bericht news: snipped-for-privacy@b9g2000pri.googlegroups.com...

I used to apply optocouplers. For real long lines I even switched to glassfiber.

petrus bitbyter

Reply to
petrus bitbyter

Not necessarily, depending on the noise sensitivity of the sensitive bits and the level of noise elsewhere. Great Big Ground Planes significantly reduce ground loop effects, but they don't eliminate them.

If it can be done, indeed.

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
Reply to
Tim Wescott

-- snip --

Caps will block DC ground noise, but will let the high frequency ground noise come sailing through unimpeded.

An ideal transformer won't provide a noise path into your board. Your goal in using a transformer would be to understand just how much coupling you can stand, and see if you can find a transformer that stays below that threshold.

The signal return path would be through that signal's local ground.

I doubt that a digital isolator would add much EMI. Yes, you'd need to power up your analog, but that could be done with a little switching supply that's transformer coupled.

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
Reply to
Tim Wescott

Capacitive coupling doesn't do much for you here, it doesn't cut the return path. You could run them through common mode chokes but with too many lines that can get old.

With a transformer your signal isn't single-ended anymore.

A good transformer will send nearly all common mode noise and ground-conducted stuff out to pasture, it's usually the way to go. The challenge is to find one that's good, cheap and most of all available. LAN transformers could be great candidates. You can get them in 2-packs,

4-packs and even with extra common mode chokes integrated.

If you have enough amplitude optocouplers are also an option and you don't have to worry about any DC restoration. One of my favorites for faster stuff is the CNW137 but that's because I work in hi-rel and medical and they come in gullwing wide body:

formatting link 

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

A good way to do this if you have control of all the spigots is to use differential signaling (RS-422 would be adequate at 5MHz), and don't connect the signal ground to the board ground. That way everything stays referenced to the same ground, but you're not providing any low impedance ground paths to your board's signal ports. You can't do this, but you can try to get the same effect.

You say your signal is "almost zero biased" -- what do you mean? Do you mean that it has just about as many ones as zeros? Can you count on this, or do you sometimes have long stretches of zeros or ones? How long?

At any rate, you should be able to feed your signal into a high speed comparator or line receiver, with a suitable local reference for your signal's 'midway' point:

VCC + ___ | .--|___|--' | | |\ o------|-\ sig out | | >----------o sig in o----------)------|+/ | |/ || | o-----||---o sig return || | .-. | | | | '-' | | === GND (created by AACircuit v1.28.6 beta 04/19/05

formatting link

This circuit assumes that your grounds are well referenced at DC, and that you can stand some high-impedance coupling. It couples any high-speed ground bounce to the comparator input to keep your noise margins up, but it somewhat isolates your board's ground from the rest of the system ground.

Depending on what you have, the comparator can me a regular high-speed comparator, or something like an RS-485 receiver (terminate to the signal return, _not_ to the board ground!), or other suitable line receiver. Just read the specifications carefully, 'cause you'll be 'misusing' just about anything called a 'line receiver' -- you want to make sure that you're maintaining isolation and providing the receiver with enough margins.

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
Reply to
Tim Wescott

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.