How to separate grounds

Thread Starter

signalflow

Joined Mar 12, 2014
50
I am trying to separate grounds for a noisy side of an optocoupler and a clean side of the optocoupler. So I would need gnd1 for clean side and gnd2 for noisy side.

I have a bridge rectifier converting the input AC voltage to DC which one of the points on the rectifier is my ground (call it gnd1). Then I have some voltage regulators. The voltage regulators are tied to this same ground (gnd1). Can I not tie one of the voltage regulator's gnd pins to gnd1 and just call this pin a new gnd (say gnd2)? But it's really just a floating pin, then, right? I'm not sure the voltage regulator would work in this case.

I just can't see how to create a separate ground. Any help is appreciated. Thanks.
 

Papabravo

Joined Feb 24, 2006
16,799
There is some ambiguity in your request. For the grounds to be isolated with an optoisolator, you need two different power supplies with no possible connection between their grounds; not on the board, not in the house not with a mouse.

To separate grounds on a board so that noise on one has to try really hard to affect the other one, you use routing and layout. Start with a common ground as close to your power supply as possible. From that single point common ground you run a trace to the "noisy components". From the single common ground you run an entirely separate ground trace to the "quiet components". You must make sure that "noisy ground" and "quiet ground" have only the single point in common as close to the power supply as is humanly possible.
 

Thread Starter

signalflow

Joined Mar 12, 2014
50
I am talking about what you state in your first paragraph (using an optoisolator with 2 separate power supplies + grounds). So basically I would need 2 bridge rectifiers + regulators, etc. I was thinking I could use 1 bridge rectifier and then 2 regulators to create the 2 power supplies, but sounds like I would need 2 rectifiers to create the 2 separate grounds. Does that sound right? What I mean is the 2 regulators would create the 2 separate Vdd lines but with only 1 bridge rectifier I would still only have 1 ground.
 

Papabravo

Joined Feb 24, 2006
16,799
Yes that is correct. There is absolutely no point in using an optoisolator to "isolate" circuits if they have a common ground at the Rectifier/Capacitor. You can of course use them for "level shifting" or "polarity inversion" if you like.
 

Thread Starter

signalflow

Joined Mar 12, 2014
50
I just don't have room for another power supply.

So you don't think having an optocoupler with the same ground but different Vcc lines will help at all? Seems like at least some of the noise would be eliminated that comes through the optocoupled lines even though some of that noise will get through the gnd.
 

John P

Joined Oct 14, 2008
1,909
If the grounds are joined, I don't think you'll gain anything with an optocoupler. The usual way to handle something like this is to separate the components into digital and analog (maybe with "high power" as a third section) and have each one on its own ground plane, with the grounds connected at only one point. Keep traces associated with each section within that section, and make the connections between them as few as possible, and just on the boundary.

I've never done this myself, though I've heard of it: you can have the boards physically separated, with fiber optic connections between them. That sounds like a pretty desperate situation!
 

Papabravo

Joined Feb 24, 2006
16,799
If the grounds are joined, I don't think you'll gain anything with an optocoupler. The usual way to handle something like this is to separate the components into digital and analog (maybe with "high power" as a third section) and have each one on its own ground plane, with the grounds connected at only one point. Keep traces associated with each section within that section, and make the connections between them as few as possible, and just on the boundary.

I've never done this myself, though I've heard of it: you can have the boards physically separated, with fiber optic connections between them. That sounds like a pretty desperate situation!
This is what I described in the 2nd paragraph of my original post. It really works great with things like stepper motors and sensitive comparators (LM311) monitoring phase currents. Trust me on this one.
 

Brownout

Joined Jan 10, 2012
2,390
You can further decouple the noisy and quiet ground by adding filtering between them. Connecting the noisy ground to the center point via a choke might help to keep the noise contained. Also, extra decoupling caps on the quiet side may help.
 

Avid0g

Joined Apr 1, 2018
21
The whole point of these exercises is to end up with a quiet DC power supply with very good regulation of the flyback oscillator and output referenced to Earth ground again.

I would like to point out to readers that using a bridge rectifier on 110 VAC Mains will have two outputs that are nowhere near earth ground. Each has a pulse DC half-sine wave, one going positive, then the other going negative.

After connecting the filter capacitors to neutral, one side is ~+187 VDC, the other is ~-187 VDC. That is with respect to neutral (which is also noisy), however, if the filter capacitors are just connected across the bridge output, then each end of the capacitors will continue to have large pulse AC waveforms at 50 or 60 Hz.

You can treat either one as a local ground for the purposes of the downstream voltage regulator circuit, like a flyback transformer oscillator primary circuit.
 

Papabravo

Joined Feb 24, 2006
16,799
Where is this alleged "flyback" oscillator. Did you forget that direct connection to the mains violates the terms of service? Using a transformer, the DC output is a good bit less than ±187 Volts Peak.
 

Avid0g

Joined Apr 1, 2018
21
Where is this alleged "flyback" oscillator. Did you forget that direct connection to the mains violates the terms of service? Using a transformer, the DC output is a good bit less than ±187 Volts Peak.
signalflow did not mention a mains stepdown transformer, so I didn't assume that there was one.

I see power supplies all the time that have the full wave rectifier bridge indirectly attached to 110v and neutral. Just a fuse, MOV, normal mode choke, and bypass capacitor before them. Then the bridge output is filtered and chopped up to drive (usually) a flyback transformer.

So signalflow was concerned about multiple signal reference frames or "grounds" (the latter which is a horrible misleading blanket term that mostly confuses) and was also concerned about how an opto-isolator comes into play. All this ties into switching power supply designs that, for the most part, use flyback transformers. So I hope you can see now why I went that way.
 

WBahn

Joined Mar 31, 2012
26,398
I am trying to separate grounds for a noisy side of an optocoupler and a clean side of the optocoupler. So I would need gnd1 for clean side and gnd2 for noisy side.

I have a bridge rectifier converting the input AC voltage to DC which one of the points on the rectifier is my ground (call it gnd1). Then I have some voltage regulators. The voltage regulators are tied to this same ground (gnd1). Can I not tie one of the voltage regulator's gnd pins to gnd1 and just call this pin a new gnd (say gnd2)? But it's really just a floating pin, then, right? I'm not sure the voltage regulator would work in this case.

I just can't see how to create a separate ground. Any help is appreciated. Thanks.
It would be extremely helpful if you provided a schematic or sketch of what you have presently and what you are considering.
 
Top