I recently encountered issues with a project involving an ESP32 and several components. Despite careful planning, I faced problems like LEDs lighting up unexpectedly due to signal interference. I’ve since learned that using a 4-layer stack-up with Signal/VCC, Ground, Ground, Signal/VCC configuration gives better results. However, I’m unsure about correctly utilizing the second ground layer.
Currently, all ground connections are linked to layer 2 with vias. What’s the best practice for utilizing the second ground layer on layer 3? Should I create vias between layers 2 and 3 while keeping all components connected to layer 2? Is it feasible to use the second ground layer on layer 3 for a few VCC traces, or does it need to be a complete plane like on layer 2?
In the ideal case you would have signals and components on the top layer using layer 2 for reference, and signals and components on the bottom layer using layer 3 as a reference. Let’s say you placed all your input circuitry on top and all your output circuitry on the bottom. That would seem nearly ideal. If the circuits are noisy then keeping the ground planes separates (maybe tying them together at one mounting hole) might be helpful. If there’s a lof of back and forth and signal loops (feedback, etc.) it might be best to tie them together closely and treat it as one ground plane.
In most cases you are best off connecting grounds together as often as practical. There might be be few circumstances where a user may want to have grounds isolated in some form but you need to understand when and why to do that. If you do not, I recommend that you do not isolate grounds. Even manufacturer datasheet recommendations have been known to sometimes get this wrong.
The way you worded your question, you suggest using blind and buried vias to achieve specific connections. I am not sure if that was your intent but keep in mind that blind and buried vias increases fabrication complexity and costs. Most common products use standard thru-vias which connect ground on all layers together and is generally the prefered condition.
Ideal conditions would always have uninterupted ground reference layers. In the real world, designers sometimes need to compromise and route a signal or power net under less than ideal conditions. Whether that choice creates an issue or not depends on the exact situation. It is not a manufacturing concern as much a potential performance issue.
Yes this is what I’m trying to achieve. I’m not worried about fab complexity as my shop can def do this but cost will increase for sure.