Power Plane split into +12V and +5V halves?

I am in the process of re-laying out an existing 2 layer board into a 4 layer board. One half of the board is I/O, micro-controller, and supporting components. The other half are high side intelligent power MOSFET's. The 2 layer board is having EMI issues with a large DC pump motor. The MOSFET's are in place to drive 12VDC solenoid valves. My plan is having the layers stacked signal/gnd/power/signal. The caveat is that I want to split the power plane into 2 halves; one being 5V under the micro-controller circuitry and 12V under the MOSFET's. I have found a lot of discussion on split gnd planes but not too much on split power planes.

Can anyone give me some direction on the ramifications of this? Also, how much separation should I have between the split?

Thanks!

Split power planes are pretty common and I woundn't worry about any issues. You should keep them apart enough to avoid and creapage or clerance issues.

As a side note I'm only taking PCB Design wise.

Hope this helps,
Chris.

PCBcomponentlibraries said:

Split power planes are pretty common and I woundn't worry about any issues. You should keep them apart enough to avoid and creapage or clerance issues.

As a side note I'm only taking PCB Design wise.

Hope this helps,
Chris.

Click to expand...

Thanks. I am now wondering if stitching capacitors might be necessary. I don't have any high speed signals crossing the gap in the split power plane so maybe I'm just being paranoid. I found this article referenced in another post. Sounds like you would need better equipment than I have access to to even know what value or where to place them.

I don't really know about stitching capacitors I'm a PCB Designer not an engineer unfortunetly!

However idealy you don't really want any tracks crossing a split GND plane as this is the return path for your signals (sorry if I'm pointing out the obvious). This becomse less important when the speed of the nets are slow.

You could only cross the power split on the top signal layer that way the tracks would be shielded from the split by the GND plane...

John, stitch caps are for Ground planes, not Power splits. Power splits MUST be completely isolated- meaning there should be absolutely NO connection - either through moats, or through beads, caps or resistors. Just place the components on your PCB such that creating different islands for 12V, -12V and 5V is possible on the PCB. And let there be one single GND plane.

Creepage and clearance do not apply at these voltages...These do not applky to SELV circuits, and on inner layers as the conductors are coverd with a class 1 insulation (the PCB material) they are less stringent.
A signals return will couple (and thus flow) in the nearest adjacent plane, thus if it is a power layer it will use this layer, if there is a split in this layer the return current has to follow a torterous path, unless you use stiching capacitors.
Stiching capacitors are for power planes please refer to pages 629-630 in "Electromagnetic Compatability Engineering" Henry Ott.
Dont get me wrong, but to design todays PCB's you require a modicum of knowledge regarding return currents and high speed design pitfalls (even the high rise times cause problems, even if the overall clock frequency is quite slow), it isn't just joining the dots any more, we as PCB designers need to educate and change with the ever increasing demands of our chosen proffesion.

Creapage and clearance:
**broken link removed**
Split Planes:
http://www.ultracad.com/articles/planesplits.pdf
**broken link removed**

Just out of interest and a bit of fun how fast do electrons travel in relationship to electrical current...

---------- Post added at 11:23 ---------- Previous post was at 11:16 ----------

Oh as to placing caps, small value, small size, minimal parasitic inductance and next to every signal that crosses (withing reason), for slower signals it isn't too critical, most boards have numerous power islands these days so it is hard to avoid crossing plane splits unless you have lots of contigous ground planes to provide the return current path...

marce said:

Just out of interest and a bit of fun how fast do electrons travel in relationship to electrical current...

Click to expand...

Going out on a limb here but isn't electrical current the flow of electrons so they would be equal?

isn't electrical current the flow of electrons so they would be equal?

Click to expand...

So speed of electrons in m/s is equal to current density*X, but what is X?

Actually you have to put in charge carrier density (e.g. 1 per metal atom, but there are many) and elementary charge, and get a very low (< µm/s) speed for common current densities in metal. It's a nice physics students homework, I suppose.

roughly 84mm per hour as an average... using mm/s = I/QeR2pi, so not very fast.
It gets interesting when you get physicist yapping about it, cos as FvM pointed out, this is based on 1 carrier per atom, if not all atoms provide a charge then velocity goes up!
I was more alluding to the fact that electrical current is x amount of electrons passing a single point in a finite amount of time, but more important is the electromagnetic waves that propogates at some velocity below the speed of light, dependant on the dielectric sourounding the wire. This IMO is important in understanding return current flow as the image we have of a train track of electrons flowing from the battery, through the switch, lighting the bulb, then flowing back to the battery gives the wrong impression, the return current is equal and opposite to the positive current and propogates in the same direction just the opposite way!!! Of course this is for DC current, for AC current the electrons just jiggle back and forth and dont realy go anywhere, so if you think about it the current is already stored in the wire (or trace).8-O
Anyway have a good weekend, and excuse the ramblings of a PCB designer:smile:

I have Ott's book on order from Amazon. The idea if only crossing the split on the top layer since the full ground plane is under the top signal layer sounds like a worthy goal.

John, this site is also worthwhile joining, its the Home of a lot of Keith Armstrong's papers and design notes:
EMCIA - EMC Industry Association, Products and Members
This page having loads of free info:
EMCIA | Free Information - EMC - EMCIA Electro-Magnetic Compliance Industry Associatio