PCB grounding

[Zozobra]

Hey folks, I've been playing around with LND150 tube emulation a la KMG and have got to the point where I think I'm more or less happy with how the PCB is laid out, although it needs a little bit of tidying before I send the gerbers off. This brings me to how best(tm) to ground things? I've tried to keep the component traces on the bottom layer as much as possible as I'm using the top layer for grounds in large pours to act as a ghetto shield. I've opted to split the ground planes to reference the filter node for the stages as appropriate. Is this a good idea or should I just go with one big pour/ground plane?

[Tony Bones]

Split ground planes can be put to good use when you also have good control over how currents move between circuit regions outside of ground.

For example, if you have a switching power supply on the PCB then all of the 'forward' current is through a single trace. In this case you put a ground plane under the entire supply and join it with the other ground plane with a single trace directly under (or over) the forward trace. Now all of the currents are moving between circuit regions through a relatively narrow channel that can be well controlled. The current through the channel on one side of the board is exactly matched by the current (in the opposite direction) on the other side of the board.

In your case you have currents moving between regions (outside of ground) through three channels: B+, -12V, and signal. If you split the ground planes then the return ground currents for all of these channels will be forced through a single trace connecting the ground planes. This will likely cause more trouble than it will cure. You might think that joining the ground planes in three places will fix that, but the ground currents won't understand your intentions and will go where they want.

In my opinion, you're better off with one uninterrupted ground plane for the entire circuit.

[Zozobra]

Thanks for the detailed response. I'm currently experimenting with explicit ground traces (where things return to their respective filter ground) with a top shield plane that is connected to an independant ground.

[loudthud]

In the USA, the square pad on an electrolytic cap is the plus terminal, but we drive on the right side of the road. Just to avoid confusion, move the + sign on the silk screen from under the cap so you can still see it when the cap is installed.

[R.G.]

Ah, grounding.

For your limited case, it will probably not matter.

At audio frequencies, what matters in grounding is what >ground offset voltages< are generated by >ground currents< flowing through >ground trace impedance<. Each active device uses some current, and when it's done with it, the current gets dumped into the resistances of the ground lead, and that generates a ground offset for that one stage.

If it's only one stage, the effect is usually to lower the gain of that stage. But when you have several stages sharing a ground offset, it amounts to a feedback connection linking the stages. Still OK if the gain of all the stages is low enough that the ground-offset feedback doesn't exceed the losses in the "feedback" and that at no frequency does the loop gain exceed unity. If it does, the thing start oscillating. If there are multiple stages with different signals sharing the same ground offset, you get cross talk and possibly oscillation.

Minimizing ground offset voltages is the key. You can do this two ways: either with tons of copper to keep the resistance low (this is ground pours) or with individual grounding traces. In most low-power circuits, either way works. Individual ground traces (also known as star grounding) is unique in that it can be proven ahead of time to always work. Ground pours and planes work, unless high current flows and unlucky ground paths are involved.

Metal plane shields work by forming a fixed, low impedance plane to intercept incoming electrical and magnetic fields. They shunt electrical fields to someplace else, and dissipate magnetic fields by making equal-and-opposite rings of current flow (eddy currents) to oppose the penetration of the field, while dissipating the incoming energy as heating in the metal resistance. Since incoming EM energy generates voltages and currents on shields, it makes sense to NOT use shield planes as ground returns. The radiated-in stuff can be added to any ground offsets, although it is reasonably rare for this to be big enough to notice in (1) low power (2) low gain (3) low frequency circuits.

Bottom line: for your stated purposes, you're probably safe either way, or both ways at the same time. As frequencies, gain, and power go up, the questions get more crucial.

[Zozobra]

In the USA, the square pad on an electrolytic cap is the plus terminal, but we drive on the right side of the road. Just to avoid confusion, move the + sign on the silk screen from under the cap so you can still see it when the cap is installed.

Noted

Ah, grounding.

For your limited case, it will probably not matter.

At audio frequencies, what matters in grounding is what >ground offset voltages< are generated by >ground currents< flowing through >ground trace impedance<. Each active device uses some current, and when it's done with it, the current gets dumped into the resistances of the ground lead, and that generates a ground offset for that one stage.

If it's only one stage, the effect is usually to lower the gain of that stage. But when you have several stages sharing a ground offset, it amounts to a feedback connection linking the stages. Still OK if the gain of all the stages is low enough that the ground-offset feedback doesn't exceed the losses in the "feedback" and that at no frequency does the loop gain exceed unity. If it does, the thing start oscillating. If there are multiple stages with different signals sharing the same ground offset, you get cross talk and possibly oscillation.

Minimizing ground offset voltages is the key. You can do this two ways: either with tons of copper to keep the resistance low (this is ground pours) or with individual grounding traces. In most low-power circuits, either way works. Individual ground traces (also known as star grounding) is unique in that it can be proven ahead of time to always work. Ground pours and planes work, unless high current flows and unlucky ground paths are involved.

Metal plane shields work by forming a fixed, low impedance plane to intercept incoming electrical and magnetic fields. They shunt electrical fields to someplace else, and dissipate magnetic fields by making equal-and-opposite rings of current flow (eddy currents) to oppose the penetration of the field, while dissipating the incoming energy as heating in the metal resistance. Since incoming EM energy generates voltages and currents on shields, it makes sense to NOT use shield planes as ground returns. The radiated-in stuff can be added to any ground offsets, although it is reasonably rare for this to be big enough to notice in (1) low power (2) low gain (3) low frequency circuits.

Bottom line: for your stated purposes, you're probably safe either way, or both ways at the same time. As frequencies, gain, and power go up, the questions get more crucial.

Thanks for the clear advice. I've managed to route a star grounding scheme with an independant shield plane without it being too messy. I have a bit more work to do on the general layout to tidy things and make it less of a headache to populate and modify and having a clearer idea about grounding is definitely going to help with that