I'm working on building a Class D Amplifier (TPA3122), and from what I understand, you want to connect the grounds of the amplifier using a star connection. On this circuit, I plan on using a DC-DC converter to power some other components (Logic IC's, LED, Isolated Bluetooth Module). So I wanted to still use a ground plane for the converter since I would need it to ensure that there isn't a large amount of inductance in my ground path.

So would it be acceptable to design a Class D amplifier, connecting the amplifier's ground using a star connection, while connecting the DC-DC converter's ground using a ground plane? Side note, I plan on using a DC-DC converter IC (haven't selected a part yet) to regulate an output voltage of 5V, and using another converter to isolate the 5V supply (CRE1S0505S3C) to power the Bluetooth module (generic bluetooth module from Amazon).

  • 1
    \$\begingroup\$ Have you read: analog.com/en/analog-dialogue/articles/… It explains why start grounding is a good strategy. It prevents the voltages resulting from large currents through ground connections, to find their way back into the sensitive inputs of your circuit. Think about where large currents flow (amplifier output, DCDC converter) and try to keep the current loops small. If you understand what happens it becomes "obvious" what the strategy should be. \$\endgroup\$ Dec 1, 2020 at 19:58
  • 1
    \$\begingroup\$ I would use multiple ground planes (one for DCDC, one for Class D amp., etc) and then star connect those groundplanes. You can use the same layer of course, just separate the individual ground planes with a copper-free area between them. See: electronics.stackexchange.com/questions/399725/… \$\endgroup\$ Dec 1, 2020 at 20:01
  • \$\begingroup\$ Following up on Bippelrekkie comment, I'd use "local battery" bypassing at the Power Switches. This provides the high currents, and the fast_switched currents, as close as possible to the switches. Have a plane under the fast switches, and ground the local_battery capacitors to that plane. \$\endgroup\$ Dec 2, 2020 at 4:16
  • \$\begingroup\$ @analogsystemsrf I'm not sure if I completely understand. Do you think that it would be better for me to use a battery? For this project, the amplifier was going to be powered from a USB C wall outlet. \$\endgroup\$ Dec 2, 2020 at 17:14
  • \$\begingroup\$ Read this: hottconsultants.com/techtips/split-gnd-plane.html. What is your ADC ENOB? \$\endgroup\$
    – user110971
    Dec 2, 2020 at 18:40

2 Answers 2


So would it be acceptable to design a Class D amplifier, connecting the amplifier's ground using a star connection, while connecting the DC-DC converter's ground using a ground plane?

Ground planes reduce common mode voltages between components by reducing the resistance and inductance of the conductors between components. In the diagram below, the higher \$Z_G\$ is the more voltage the loads \$R_{L1}\$ and \$R_{L2}\$ will see. Now if you suppose that \$R_{L1}\$ or \$R_{L2}\$ are transistors, and \$Z_G\$ is a trace that runs back to the power supply. The first problem is if the current changes through the transistors, it will change the 'ground' voltage for both.

Lets suppose the current is 100mA through a 1" 10mil 0.5oz trace (which will have ~50mΩ of resistance. The voltage (V=IR) generated from the resistance of the trace would be 5mV!

Now lets suppose we use a ground plane instead which would have roughly 0.3mΩ of resistance in one inch, we would only have a 30uV voltage generate from the ground plane resistance with a 100mA current.

It gets more complicated than that with ground planes because the current spreads out in two dimensions (and currents should be controlled in two dimensions for very low level designs), but the advantage of a ground plane having less resistance can be seen. In almost all designs using a ground plane is better than running traces.

If I were designing a two plane design and I couldn't have a ground plane and only run traces, I would calculate the trace resistance and simulate the design with the 'parasitic' resistance of the traces to see how it behaved.

enter image description here
Source: (pg 129) Electromagnetic Compatibility Engineering. Author(s):. Henry W. Ott.


Ground planes reduce the influence of current spikes on other components and provide some shielding against RF interference. They are poison for signal routing in amplifiers since 50Hz/60Hz magnetic fields are omnipresent to some degree in every building and their EM wavelength is such that you cannot shield them with electrically conductive materials but only with hefty amounts of mu-metal. This means that your reference grounds should be routed closely paralleling the signal paths. As soon as you have more than a single channel and unbalanced inputs grounded with a common front plate, you will not be able to route stuff without needing to cut up and mix reference grounds through resistors limiting ground loop effects (ballpark size 10 ohms, with the ratios of resistors from ground references arriving at the + terminal of the mixing opamp matching the ratios of the respective signals including the output signal at the - terminal).

So essentially you want a ground plane topology for all bypass capacitors leading to the center tap of your transformer that is the ultimate balancing current source and sink, but you want a star topology for your reference grounds, with the violations of the star topology explicitly accounted for by a separate ground compensation network reflecting the signal topology.

This appears to roughly reflect what you are planning for (plane for power, star for signals), except that you don't appear to have a strategy in place in case you need to violate the star technology that is fully workable mostly when working with a single channel and/or balanced inputs.


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.