I have a few questions about how grounding on a PCB and to an enclosure should be done when using multiple supplies. I have read a number of similar questions/answers from multiple sources with the most definitive being
Should chassis ground be attached to digital ground?
However this doesn't quite answer my questions for this specific case.
I have a single PCB board that has two circuits: one (denoted TC) powered by a +/- 5 V split supply, and the other (denoted CC) by a +/- 15 V supply. Both circuits are connected to a 14-pin butterfly package and everything is to be housed inside a metal enclosure. Only one pin of the butterfly package is connected to the case and only to the case. For heatsinking reasons, the case is mounted to the metal enclosure. A basic layout is as shown:
The two circuits are separated, so in principle could have separate grounds. Due to the pin configuration of the butterfly package, current (~0.5-1 A) return for TC is on the CC side of the board. A separate trace could easily be routed for the return, but the lowest impedance return would be along the edge of the board anyway.
The CC circuit includes a DAC (AD5790) that is programmed by an on-board Arduino Nano BLE. The Arduino board is independently powered and the DAC has its own 3.3 V digital supply (derived from the +15 V). The intention is that the Arduino is only used intermittently and can be powered down and disconnected when not in use. The DAC placement and routing is such that the digital part is separate from the analog parts of the board. The Arduino has its own board but can be hard soldered to a PCB. My intention would be to connect its GND to the GND plane of the PCB. I don’t think there is any other way to (easily) attach the GND e.g. through the USB connection.
Powering the TC from the same 15 V split supply is not really an option as it will result in unnecessary power dissipation (5-10 W).
My questions are:
- The datasheet for the DAC suggests a separate GND for DGND and AGND and only connect it at one point to “avoid ground loops”. Given that I routed the DAC signals on the top layer immediately above the ground plane and away from the analog section, is there any reason to separate the GNDs? One argument I read was that signals will propagate such that return currents will remain under the signal traces and such a separation is unnecessary and more likely to cause problems. If they should be separate, where would the single connection point be?
- Should TC and CC have a separate GND planes? Given they a physically separated anyway this seems a little redundant but they are connected to two different supplies.
- How should the supplies be brought into the enclosure? I considered screw-in feedthrough capacitors for the supplies and turrets for the ground lines with the hope of killing any high frequency on the supply lines but wondered if these would be better isolated from the enclosure given the two separate supplies.
- How should the enclosure be grounded? From the link above it seemed that the “best” solution would be to use mounting holes for the PCB that are plated and connected to the GND plane(s) to tie the PCB to the enclosure, but I assume this depends on how everything else is (should be) done.
Regarding the supplies, any advice on the cables would be appreciated. The enclosure will sit probably 5 m away from the supplies and I am concerned about what might get picked up from everything else in the lab (often in the range of a few MHz to a few hundred MHz).
