Hi. I have designed a 4-layer implementation of an Analog Devices LT8711 Zeta Converter.
I have the gerbers for an evaluation board. The designer has connected the signal and power grounds on the top layer and bottom layers and also on an internal layer. The connections are at the top of the 20-pin TSSOP footprint in this picture
I have used net-tie only once to make RFID antenna (shorts with copper its both ends).
I think tere is no way to define net-tie at inner layers, but it should be possible to define THT net tie for all your THT pads under the IC. May be such one net-tie will solve your problem.
But as the signal and power grounds are probably separated in area you just can make them both being the same net (so no net tie problem to be solved) and define their shape separately with overwrite one over another under IC. It can also be one GND zone with right shape.
KiCad still has a limitation to place footprints only on the front and back layers, and net ties are implemented as footprints (with pads, so you can connect to them). I once saw a hack in which the PCB file was edited in a text editor to place footprints on an internal layer, and apparently this works and you can then move the footprint in KiCad and it stays on that layer.
But I have some doubts about the usefulness of all this. As far as I know this design method is mostly a stubborn artifact from days gone by. As far as I know, more GND is almost always more betterer. And then combine it with physical distance between “noisy” parts (SMPS) and “sensitive” sections (ADC frontend). All ripple current with frequency content higher then a few kHz will always follow the path of least impedance, and not the path of least DC resistance. If at all possible, the currents in a continuous GND plane will always be directly under the track, because this is the smallest loop inductance, and thus lowest impedance. If you make slots in your GND plane, then you force the GND currents around those slots. Thus higher impedance, and you generate more noise, and you hope to mitigate the effects of this higher noise by using more slots in your GND plane. It’s yucky.
To make things easier, we can readily debunk all of the foregoing and say they are not true. But, perhaps one of the most important takeaways is that you should NEVER, EVER split ground planes. If you do, you will destroy the integrity of your PDS. Splitting Planes—The Good, The Bad And The Ugly | PCB Design Blog | Altium
I just had a bright moment. These days 4-layer PCB’s are common, and assigning one or both inner layers to GND (Prepreg is much thinner then the PCB core, and thus better coupling and less noise) is much easier then the old days, when most designs were on two layer PCB’s. If you only have two layers, then there usually is no room for a full GND plane and some sort of compromise is the result.
Also: How to achieve proper grounding youtube video from Rick Hartley is (over two hours long) worth watching.
I make net-tie footprints with PTH holes and then inner layers are connected. Just a couple of overlapping oval or rectangular pads for example. Throws drc errors that I ignore (could write a rule I suppose).
I’ve followed the layout guidelines closely but now combined the 2 grounds. Heavy and fast switching currents will be localised to the top left.
The controller is in a “quieter” area of the PCB.
Top layer is Vin, Vout, GND and high-speed switching path (MOSFET drains etc). Second layer is the same. 3rd layer is MOSFET drains etc, Vin and Vout. Bottom layer is same as top and 2nd but with a lot of tracks from the controller up to the MOSFETs and sense resistors. There are some bypass capacitors on the bottom too to save space on the top.
Thanks, that is nice! I was not able to get your file running, but I tweaked an smd net-tie from the kicad library to make it 4-layer and it works well. I will post it in the other thread I started on net-ties.