Bipolar Supply, Unbalanced Load

[PlasticIsGood]

Originally Posted by DF96

His wiring is correct: the dirty stuff is at the CT end and then he grounds the clean end at the other end of the copper.

I don't see why the caps are connected to all four corners in the second diagram, if that's a bus. If it's a star, then I don't see the difference between it and the first diagram.

[DF96]

It is intermediate between a bus and a star. Think of it as a distributed bus. It could easily be a bus, if the copper were more narrow. As in all grounding, think about where the currents flow. In this case you have charging pulses between the CT and the left (reservoir) cap grounds, then ripple current from the left cap grounds to the right (smoother) cap grounds, then nice pure DC plus clean signals out of the other end to the circuit and the speakers.

[cfsatdiyA]

I came across this thread trying to address a similar problem. Hopefully the attached schematic and simulation results add a little more insight to anyone that may still be following this thread or comes across it.

As I feel demonstrated by the discussion, so many struggle with these concepts (myself included). I feel the basic problem is identifying the parasitics. Star grounds can cause as much problems as they fix if the basics of the problem is not understood. However, I do have some fundamental rules I like to follow.

A shield is not a ground return. Make sure you are not confusing a circuit ground with shielding although a circuit ground can provide shielding (so can the positive rail). Account for your return currents. On the later, if it has a goes outa, it needs a goes inta. This is where I see the fundamental problem lies for this situation.

The image of the schematic I have attached includes a parasitic resistance, R9, from the CT of the transformer to the filter capacitors (the xfrm secondary output is represented by two opposed voltage sources V1 and V2). R10 represents the parasitic resistance from the capacitors to the power supply ground. The loads, R8 and R5, hopefully represent the situation that the original poster described. As can be seen by the simulation the unbalanced load causes significant current pulses I(R9) on the CT of the transformer. This is the charging of the filter capacitors. They only charge when the sine wave voltage from the transformer exceeds the diode drops of the rectifier. In addition the difference in the return currents from the outputs of the regulator cause a DC offset current I(R10) on the mutual ground of the two regulator circuits. However, it does not contain the pulsing current because the regulators have effectively filtered it out due to their PSRR and the function of the filter capacitors.

Hopefully this schematic shows that a star ground does not solve the problem if you put the star in the wrong place, like the output of the bipolar power supply. See second schematic. This will induce 8mVp-p of noise on the ground even with just 0.01ohms of parasitic resistance back to the filter capacitors. Connecting it back to the previous circuit but keeping R11 this reduces to below 1uVp-p of ripple on the ground. There is a 1mV DC offset though. Consider this though, a .150” wide 1oz copper trace has ~.001ohms/inch. Ground planes cover a lot of sins.

When the OP built two separate isolated supplies he directly solved his ground current problem. He had a goes outa coupled to its goes inta for EACH supply. With a bipolar supply sharing a command ground the currents add and subtract. You will see the difference and the parasitics are there to show you they exist.

Learned something new.

CFS