0
\$\begingroup\$

I have the following layer stack up: Layer 1: Components and traces Layer 2: Dedicated ground plane Layer 3: Power plane Layer 4: Additional traces as needed

I have read (I'm a beginner at circuit design) that it is best that traces have a ground plane either on the layer immediately below or above said traces for return path. However, given I have some traces on layer 4, the layer immediately above these traces would be my power plane, not my ground plane. Would it then make sense to place a relatively small ground plane island within the greater power plane (layer 3) that corresponds to the location of the traces on layer 4 such that there is a ground plane directly above these traces? And then stitch this ground plane island to my dedicated ground plane (layer 2) using vias? Within the attached image, you will find that I have attempted to create such a small island directly beneath the feedback traces for my buck converter (stitching vias not yet added). Would this be beneficial? Is it not required for these short traces (app. 15mm in length)? I've already included a ground plane for signal trace return paths within the power plane layer (seen in the second image), however this seems more straightforward as my power plane doesn't need to be in the top portion of my PCB in the first place. Thank you.

enter image description here

enter image description here

\$\endgroup\$
7
  • 1
    \$\begingroup\$ What exactly is on your island that needs this special treatment? Looks like 3 resistors and 2 capacitors. This seems like a terrible reason to break the plane. \$\endgroup\$
    – MOSFET
    Commented Jul 20 at 2:31
  • 1
    \$\begingroup\$ Your feedback node is sampling the output of a capacitor and so should not have high frequency signals that would need a low inductance return path provided by a ground plane. \$\endgroup\$ Commented Jul 20 at 2:54
  • 1
    \$\begingroup\$ You might want to consider using thicker tracks where possible and where higher currents are flowing. The routing looks strange - was an autorouter used? Look at production pcbs to see how they generally route tracks. \$\endgroup\$
    – Kartman
    Commented Jul 20 at 4:19
  • 1
    \$\begingroup\$ Why the enormous courtyards around some components? \$\endgroup\$
    – Hearth
    Commented Jul 20 at 5:46
  • \$\begingroup\$ @user1850479 Ah, I see. I was unaware that it was only high frequency signals that would need the low inductance return path provided by a ground plane. Judging from what others have answered, it now seems clear to me that cutting my power plane would be very not good! Thank you for your help and your time. \$\endgroup\$
    – NickRand
    Commented Jul 20 at 23:59

2 Answers 2

2
\$\begingroup\$

This is going to be a "it depends" situation. Specifically, what the circuit is actually doing: the currents involved, the frequencies, analog and digital coupling.

First rule of thumb:

Don't break the power plane unless you can objectively justify why you are doing that

Without knowing much about your circuit, what I'm seeing, is that you are going to create a parasitic antennae. This is bad for two reasons: one it will radiate interference. This will be fun during compliance testing. And two, it will also be susceptible to external interference.

Fortunately, fixing one of these problems fixes both at the same time. The whole philosophies and theories behind ground planes is rooted in high frequency current return paths - AC current takes the path of least impendence NOT necessarily the path of least resistance. This is the whole idea of having a ground path (the ground plane) under the trace - it minimizes the loop the current takes thus its inductance. Lower inductance is therefore lower impedance at AC. And Impendence is basically the resistance to (high-frequency) AC.

Having said all that, just be mindful of the AC return paths and you should be good. With the little info I know about your circuit, and my experience, can get this done with a solid plane. Which is generally ideal.

Reasons for breaking the plane: You need high voltage isolation (creepage/clearance). Or, you need to isolate sensitive circuits from ground bounce. Both cases require you know what you are doing, but the latter really needs thought put into before you make things worse. Like way worse.

\$\endgroup\$
1
  • \$\begingroup\$ Alright, that makes sense. So, avoid breaking the plane unless I have a very specific and well-thought-out reason to do so (which, as I'm sure you can guess, I have no very specific and well-thought-out reason to break the plane). I appreciate you taking the time to answer my question - thank you very much! \$\endgroup\$
    – NickRand
    Commented Jul 21 at 0:15
2
\$\begingroup\$

This is in reference to your previous question? What should I do when clearance boundaries overlap with SMD pins?

Feedback traces don't need any special treatment. Definitely don't cut up a plane (a far worse sin, not something to be done lightly and unknowingly) when signal bandwidth is low -- these will be MHz if that; even loose wires will suffice.

Likely more important is avoiding direct coupling from the switching node to these traces. This is easily avoided by putting them (DW and FBs) in different locations, or on opposite sides. As-is, you have both, and I gather, a reference plane of VOUT, which is likely quiet enough not to be a problem.

\$\endgroup\$
1
  • \$\begingroup\$ Hello Mr. Williams! It is now clear to me that I must avoid cutting up a plane unless I have a very specific reason to do so, and I now see that the potential reasoning I gave in the original answer is definitely not a reason to cut up a plane. Thank you very much for your help and time - again! \$\endgroup\$
    – NickRand
    Commented Jul 21 at 0:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.