PCB Unconnected Copper Pour - Bad Idea?

[abbey road d enfer]

It's difficult to think of a scenario were it would have a negative effect.

Probably because you haven't seen as much bad design than me.

Abbey road d enfer's example (which I assume is trying to demonstrate some kind of loop that might have currents induced around it)

Wrong ASSumption.

[/quote] is not particularly realistic since all of the ground leads would stitch together the two planes [/quote] Have I mentioned two planes? I've just shown an example of how a copper pour actually connects points that souldn't be.

and break up the loop making it (them) insignificant.

No loop involved. Think resistance and current circulation.

And the drawing is not representative of how the parts would be laid out.

Really? How do you know that?

It would have to be all surface mount spread out over a large area with no vias.

Again, I haven't mentioned dual copper pours. The problem I hint exists in single-sided layouts.

And something would have to be throwing some serious current around.

But there is! Current in the rectifiers and associated smoothing caps is a major source of pollution introduced in the audio reference ("ground").

I think it would challenging to deliberately design a board that exhibited negative effects from a ground plane loop like that.

Who said it is deliberate? AFAIK, autorouters don't have the intelligence to organize ground circulation hierarchically when all "grounds" are labelled identically. An autorouter would gladly connect the xfmr center tap to an input ground if it "thinks" it's convenient.

 

[diggy fresh]

Thanks guys!  Thanks alot Squarewave, very clear and informative.  I try to learn about that stuff but most info on these topics deal mostly with RF, that helps!

There, for example.

Ah! I can clearly see the "loop".  Thanks!
I guess in that case, it's best practice to separate the psu ground from the rest and connect the "post regulator ground" to the audio circuit ground plane, pours only at one spot. (And keep the center tap and input caps ground separate with very fat and short traces.)
But appart from the problem on your drawing, what other negative effects could there be? (In simple audio circuits, no rf, no power amps, no digital).

  I am generally more worried about added capacitance from an op amp - input to ground that can add lag to the NF signal and diminish stability.

JR

Thanks John, That's the kind of mechanism i was worried about but don't know enough about.

I usualy try to surround the input signal with ground, pours side-to-side and also on the top layer with a via close to " shield" it and minimise the loop as in RF techniques, Right up to the input pin if possible.

I guess this creates a lot of capacitance to ground (and filter RF a bit too?)

With what you said, I'm guessing in theory the capacitance to ground on the negative feedback path of an opamp would reduce the NF on some very high frequency and could cause oscillation ? 

I'm trying to understand, what would be a solution to this?  leave space around the nf paths? Would that problem only be with very high bandwidth opamp?

I find this stuff very interesting.  thanks guys

John

 

[Bo Deadly]

Probably because you haven't seen as much bad design than me.
Wrong ASSumption.

Have I mentioned two planes? I've just shown an example of how a copper pour actually connects points that souldn't be.
No loop involved. Think resistance and current circulation.

And I'm supposed to understand that from that graphic that has basically no description and has almost nothing to do with the question that was asked which was if the upper / extra / second ground plane could have a neg effect?

Whatever. Anyways ...

So you're saying that if amp inputs are electrically close to the transformer center tap but the filter caps where on the other side of the board for some super dumb reason then you might get some mains hum because the ground plane resistance between the filter cap and center tap is greater than between the amp input and center tap? Meaning it might be illustrated by a "resistor" between the center tap and filter caps that the signal ground doesn't have to go through?

 

[ruffrecords]

Maybe when you deal with a very dense PCB containing several signals that may or may not be mixed, such as a single PCB mixer, you'll see the evidence.

You are right. It is all to easy to answer a question with only your own experience in mind. Even easier to forgaet that even the relatively narrow field of analogue pro audio electronics span a surprising range of requirements.

Anyway, the subject is not about copper pour: yes or no, it is about floating or not, and I can't imagine a case where a floating pour gives any advantage over non-floating.

I agree that is what the OP asked but I thought my additional question was relevant.

Cheers

Ian

 

[john12ax7]

I suppose the relevant follow up question is then what is better,  grounded top copper pour vs no top copper pour.

From my experience,  there is greater dimensional stability when copper is more evenly distributed.  No top copper pour can lead to warpage,  but then again audio circuits are in fairly benign thermal environments,  so may not be issue.

 

[abbey road d enfer]

And I'm supposed to understand that from that graphic that has basically no description and has almost nothing to do with the question that was asked which was if the upper / extra / second ground plane could have a neg effect?

Adding a second ground plane may very well create the kind of connection that must be avoided.

So you're saying that if amp inputs are electrically close to the transformer center tap but the filter caps where on the other side of the board for some super dumb reason then you might get some mains hum because the ground plane resistance between the filter cap and center tap is greater than between the amp input and center tap? Meaning it might be illustrated by a "resistor" between the center tap and filter caps that the signal ground doesn't have to go through?

It's not a matter of greater or smaller really; it's about isolating currents.
"diggy fresh" has correctly understood what I meant.

 

[Bo Deadly]

Adding a second ground plane may very well create the kind of connection that must be avoided.
It's not a matter of greater or smaller really; it's about isolating currents.
"diggy fresh" has correctly understood what I meant.

It would be a lot better if you just describe the actual mechanism instead of speaking in veiled statements like "for example:" and "Think resistance and current circulation." and "it's about isolating currents". It's like you're playing Pictionary. Is it a mouse playing a trumpet?

 

[JohnRoberts]

Thanks John, That's the kind of mechanism i was worried about but don't know enough about.

I usualy try to surround the input signal with ground, pours side-to-side and also on the top layer with a via close to " shield" it and minimise the loop as in RF techniques, Right up to the input pin if possible.

I guess this creates a lot of capacitance to ground (and filter RF a bit too?)

With what you said, I'm guessing in theory the capacitance to ground on the negative feedback path of an opamp would reduce the NF on some very high frequency and could cause oscillation ? 

This is getting a little esoteric and too much information for many but since you asked... I will try to keep it short.

In negative feedback stability concerns the NF in fact needs to be attenuated below unity gain by the time phase shift/time delay accumulates to 180' turning the negative feedback into positive feedback, and oscillation. A RC pole formed by the NF resistor and C to ground at - input is more problematic for adding phase shift (lag) in the wrong direction making the circuit less stable.  A simple remedy is adding some lead capacitance across the feedback R, but you can't use too much feedback capacitance if the op amp is not unity gain stable.  I have seen very small feedback caps used to neutralize stray capacitance in non unity gain stable op amps. 

I'm trying to understand, what would be a solution to this?  leave space around the nf paths? Would that problem only be with very high bandwidth opamp?

I find this stuff very interesting.  thanks guys

John

You can waste a lot of time micro-analyzing minute details like this. Guys like Abbey spent way too many hours doing exactly that.

@squarewave, this is not a political argument, take a breath and try to understand what Abbey is sharing.

JR

 

[john12ax7]

A common technique for high speed design is to remove copper around sensitive nodes if the capacitance is an issue.  You can see examples of what to do in datasheets for sensitive high speed parts.

If using common slower audio opamps it's not a big concern.  Good grounding and power supply decoupling should be enough.

 

[CJ]

just do the opposite of what all the theorizing says and it will work  perfect.

interesting note- the Langevin AM16 board uses parallel traces to create capacitance.

edit: by parallel i mean one over the top of the other, which is parallel in that there is more than one way to be parallel in 3 space.

 

[abbey road d enfer]

nteresting note- the Langevin AM16 board uses parallel traces to create capacitance.

edit: by parallel i mean one over the top of the other, which is parallel in that there is more than one way to be parallel in 3 space.

Using parallel traces for balanced signals makes a lot of sense, since whatever crap is picked by one is also picked up by the other, but phasepolarity-reversed

 

[Newmarket]

  Who said it is deliberate? AFAIK, autorouters don't have the intelligence to organize ground circulation hierarchically when all "grounds" are labelled identically. An autorouter would gladly connect the xfmr center tap to an input ground if it "thinks" it's convenient.

True. That's why some ECAD systems allow you to 'Bridge' differently named nets but it usually comes along with a warning that automatic rules checking may be compromised.
For myself I generally find it better to use differently named nets and link them with a zero ohm resistor/link as appropriate.
and tbh I almost never autoroute.

 

[moamps]

Using parallel traces for balanced signals makes a lot of sense, since whatever crap is picked by one is also picked up by the other, but phasepolarity-reversed

Parallel traces for balanced signals makes sense because the unwanted signal picked by traces will be equal in level and phase.

 

[JohnRoberts]

I've shared this before but back when I was stuck using single sided PCB technology my senior design engineer used a trick where he bracketed a signal trace with two return traces one one either side, this effectively collapsed the loop area between them all to zero...

With multilayer boards you could probably sandwich traces between shielding above and below.

JR

PS: Sorry TMI again

 

[john12ax7]

With multilayer boards you could probably sandwich traces between shielding above and below.

In RF it's called a stripline,  a classic transmission line technique that's been around for a long time.  Can be effective,  but also potentially cause more problems than it solves.

 

[ruffrecords]

In RF it's called a stripline,  a classic transmission line technique that's been around for a long time.  Can be effective,  but also potentially cause more problems than it solves.

Stripline was co-invented by one of my lecturers, Pete Johns, at Nottingham university. In his spare time, Pete played church organs. I recorded him playing Bach's Tooccata and Fugue on the organ in the Portland building at the university. This was later broadcast on the student weekly radio programme 'Campus' on BBC radio Nottingham in about 1971. It was the first recording I made ever to be broadcast. I still have the tape.

Cheers

Ian

 

[abbey road d enfer]

Stripline was co-invented by one of my lecturers, Pete Johns, at Nottingham university. In his spare time, Pete played church organs. I recorded him playing Bach's Tooccata and Fugue on the organ in the Portland building at the university. This was later broadcast on the student weekly radio programme 'Campus' on BBC radio Nottingham in about 1971. It was the first recording I made ever to be broadcast. I still have the tape.

Cheers

Ian

Ah! The importance of the organ in the British audio community. Graham Blyth, co-founder and head designer of Soundcraft is a brilliant organ player, and SSL was founded with the purpose of creating switching systems for pipe organs.

 

[JohnRoberts]

When I got out of the army in the early 70s MIT Instrumentation Lab hired me back as a technician, as was their responsibility, but they did not give me my old job back working on a navy rescue submarine, instead I got hired into a different group working on inertial navigation for missiles and the like  ??? .  Not my idea of an admirable mission... but back on topic, one of my projects (before I escaped into a different industry as quickly as I could) was a crude RF waveguide. I constructed a physical structure by soldering copper clad  PCB stock into a box, tube, or whatever.

JR

 

[Newmarket]

Ah! The importance of the organ in the British audio community. Graham Blyth, co-founder and head designer of Soundcraft is a brilliant organ player, and SSL was founded with the purpose of creating switching systems for pipe organs.

Indeed. And when i was doing some work for Makin Organs back in the 90s the technical director was Hugh Banton (ex of Van der Graff Generator)