Loopy Ground Loops

A while back, I posed a question about using flood fill (AKA copper pours). I've been reading a lot about ground loops lately which brought me back to that original question.

LED scroll ground plane Some people suggest segmenting your ground plane between analog and digital sections. Some people suggest segmenting the ground plane for individual critical ground return paths. The follow on to my original question is: On non-exotic designs does segmenting ground planes really help? There's actually two questions, with the second being: At what clock speed does it make sense to start worrying about issues caused by ground return paths / ground loops? There are probably more questions. Those are just the two rattling around in my head at the moment.

Interestingly, though, when I wrote the original post, there didn't seem to be a clear "most common" between pour and no pour PCBs. Today, I'd have to say that the majority of designs we see here at Screaming Circuits do use flood-fill ground planes, either internal or external.

Duane Benson
You can solve ground and noise problems by just not hooking up power

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c008a53ef01538f297e8b970b

Listed below are links to weblogs that reference Loopy Ground Loops:

Comments

I saw a suggestion once to carefully segment your board and then place 1206 resistors, closely-spaced, along all the segment boundarys so you could 'stitch' the board back together with 0 Ohm links to evaluate the difference.

I actually tried this on a design that incorporated a low-power RF receiver, sensitive analog circuitry, a microcontroller and two 36V 10A brushed motor drives.
I was very careful to ensure no current return paths crossed segment boundaries, all inter-segment comms was isolated etc...

The performance with the segmented design wasn't great, the noise from the drive was getting in everything. But, when I stitched the board togther I got a large increase in performance.

I was a bit supprised how much difference there was. I wasn't really expecting any, or at least just a slight improvement in the segmented version.

So, it definitly put me off segmented ground planes. I did look carefully and couldn't find a ground-loop or bad current return path...either way it didn't work well!

I think more than clock speed, you need to look at edge rates, no? Clearly fast clock speeds require fast edge rates, but if you have a chip that is capable of running really fast, yet you clock it slowly, it may not help you that much with EMI, since the edge rates are what contain the higher harmonics.

My past experience, and advice from other knowledgeable engineers is: do not segment your ground plane. Just work really hard to maintain a good, solid ground plane. So far, this has been reasonably good advice for my designs (which always contain pretty low level audio signals). That does mean ignoring all the A/D converter datasheets that tell you to segment the plane.

If you do segmenet your plane, you need to be absolutely sure that you could use a pair of scissors (well a dremel tool anyway), and cut the board along the GND segmentation and not cut any traces or planes. Any traces that do cross the boundary should be treated as if they are going off board -- they may need EMI filters or other devices to prevent the high frequencies from getting on to your analog section and destroying your EMI performance, and perhaps affecting your analog performance.

Segmenting power planes is often required, so if you do, be sure that any signals that cross a boundary has a proper GND reference -- don't count on your power plane being the reference.

Post a comment

If you have a TypeKey or TypePad account, please Sign In.

« Package Variants | Main | Random Via-In-Pad Myth #5 »