Dual Supply Pcb power plane?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

JAY X

Well-known member
Joined
Jan 9, 2009
Messages
683
Hi All,

I'm currently designing an analog audio two layer pcb, with power and signals on top layer, and ground plane in bottom.

I use dual opamps like ne5532 lm833p, and  two 24v dpdt relays.  Now i'm laying out the power traces, and sometimes it gets a bit difficult to trace supply lines for every opamp. I still have to try 4 layer board design, and related to 4 layer i have a question.:

¿ is it possible to design a DUAL supply power plane?  usually searching on the net i see SINGLE supply power planes, as used for digital stuff... but not dual.


Any idea or comment is wellcome!
thank you very much,

JAY X
 
JAY X said:
Hi All,

I'm currently designing an analog audio two layer pcb, with power and signals on top layer, and ground plane in bottom.
Why would you want to impose on yourself such a constraint?
I use dual opamps like ne5532 lm833p, and  two 24v dpdt relays.  Now i'm laying out the power traces, and sometimes it gets a bit difficult to trace supply lines for every opamp.
That's why double-sided PCB's were created. Makes everything easier.
I still have to try 4 layer board design, and related to 4 layer i have a question.:
I have never seen a case where 4 layer was undoubtedly justified for a pure analog board. Mixed-signal is a different case.
¿ is it possible to design a DUAL supply power plane?  usually searching on the net i see SINGLE supply power planes, as used for digital stuff... but not dual.
It's certainly possible with some ingenuity, but I think a good reason why you can't find it on the web is because it's utterly useless. As much as ground planes are useful, power planes in audio are just a waste of copper, unless you're dealing with kilowatts.
 
Hi Abbey!

Well, so i stick with two layer...Ok. So i understand that is Ok to place power tracks on the Bottom layer, or alternate top and bottom layers using vías.

My objective is not to break the ground plane too much, with vertical traces. I try to follow the criteria: " ground follows signal", and as far i have seen/read,  power and ground should follow the same path.. to avoid ground loops... so power and ground follow the signal...in theory...

Besides there is the ground plane: With it, and thermal pads ¿can i avoid the use of a ground tree or star ground?

Thank you for your help!
JAY X

 
JAY X said:
My objective is not to break the ground plane too much, with vertical traces. 
Do you have "copper pour" ? It helps a lot.
I try to follow the criteria: " ground follows signal",
I'm 100% with you.
and as far i have seen/read,  power and ground should follow the same path..
not necessarily
to avoid ground loops...
You won't create ground loops by running power differently than ground; you may however create current loops. These can be easily managed by placing large value decoupling caps (electrolytics in parallels with ceramics) near the stages that have large return currents.
Besides there is the ground plane: With it, and thermal pads ¿can i avoid the use of a ground tree or star ground?
You don't want a star ground (look at my sig). You must create your hierarchical ground (that's the accepted name for "ground follows signal") and then manage the polygon that defines the copper pour in such a way that your ground plane is just your swollen original ground track. The polygon should start at the rails reference point (either the 0V connection if power is separate - make sure the common of decoupling caps is connected via a solid track to this point - or the regulators reference if power is on board).
 
Hi Abbey!,

Thank you for your answers, they are very clarifying, things that i didn't know!!.

Now it is time to take these into practice!! :)

JAY X
 
Hi Abbey Road!

JAY X said:
You must create your hierarchical ground (that's the accepted name for "ground follows signal") and then manage the polygon that defines the copper pour in such a way that your ground plane is just your swollen original ground track. The polygon should start at the rails reference point (either the 0V connection if power is separate - make sure the common of decoupling caps is connected via a solid track to this point - or the regulators reference if power is on board).

I was thinking about it yesterday...let's see if i understood it!: :)

So, if i trace a ground track... an "L" or an inverted "T" under an opamp,  following the inputs and outputs of the opamp.., then i could fill a polygon that follows the ground trace, that ends "swallowed" by the borders of the ground polygon.

If this is the way of doing it, then, the original "ground tree" that brings together the grounds of all the circuits, will be make part of the border of the polygon that defines the ground plane. ¿is that right?

To make this i have to use the "copper pour" and "polygon fill" functions of my pcb software. It is a combination of both.

JAY X
 
Hi,

Still working on it!.. if it was as simple as flooding ground..placing thermal pads.. an let the return current flow...!! ;)
And now with the ground plane concept.. i can't "see" where the currents WILL flow... :-\ ¿is there a method to see/expect current circulation path?

Noise/EMC wise, i will stitch both layers around the perimeter of the board. ¿Is it bad for crosstalk to have parallel traces on opposite layers? (2 layer pcb) for example: balanced summing inputs: + input on top layer an - input on the bottom, parallel
¿Or better is that each trace has ground plane opposite?

JAY X

 
JAY X said:
¿is there a method to see/expect current circulation path?
No magic method. Just hard work. You have to think where a voltage comes from, where it goes and where the current flowing into the load returns to ground.
Noise/EMC wise, i will stitch both layers around the perimeter of the board. ¿Is it bad for crosstalk to have parallel traces on opposite layers?
It is indeed bad for x-talk: one should isolate as much as possible two signals that should be separated as much as can be (like in "stereo")
  for example: balanced summing inputs: + input on top layer an - input on the bottom, parallel
This cannot be qualified as x-talk. These are balanced signals. Their paths should be  as close as possible; there is nothing better than running them an two different layers. In addition, the tracks are close-coupled, with makes their mutual inductance profitable for common-mode rejection this only matters at MHz).
¿Or better is that each trace has ground plane opposite?
No. But in the end the difference is lame.
 
Hi Abbey!

Thanks for your answers, they place light in these sometimes grey matters!!

JAY X
 
Back
Top