Proper ground in preamp using centre-tapped power trafo

[dasbin]

I built a couple of Fabio's 312 boards and have them nicely in a rack case, and all seems well (well, mostly... check the 312 thread her in the lab for more). I'm wondering, though, what the ideal solution to for grounding is. Right now I have a star-ground on the chassis that EVERYTHING goes to:

- Power trafo 0V Centre-tap
- IEC Earth
- Ground reference for the 312 cards
- Input and output cable shields (through chassis)

I have a nagging feeling that this isn't the right way to do it, and that my 312 card grounds should be seperate. The obvious way for this to happen is to remove the power trafo centre-tap from the star-ground and connect it only to the cards.
Unfortunately though, as soon as I do this, my power trafo stops working and doesn't give any voltage on the outside taps. I've had this problem in the past with every centre-tapped transformer I've ever used... unless the centre-tap is connected directly to IEC ground, the trafo doesn't work. Why is this? I thought it was DESIGNED to be an isolated ground.

 

[PRR]

> I have a nagging feeling that this isn't the right way

Why?

> remove the power trafo centre-tap from the star-ground and connect it only to the cards.

NO!

Power transformer CT goes straight to power capacitors CT. Then connect that point to your star ground.

 

[dasbin]

[quote author="PRR"]
Why? [/quote]
Recently did some reading that suggested that the ground being used for shield on balanced cable interconnects shouldn't be shared with the audio ground because any noise drained by one will modulate the ground of the other.

NO!

Power transformer CT goes straight to power capacitors CT. Then connect that point to your star ground.

Sorry, yeah, I was actually thinking of my PSU ground not the cards, not sure why I wrote that.
Why does that CT also have to be connected to star ground though? I still haven't figured out why the floating artificial ground of the CT doesn't seem to work with any power trafos I've tried, without being tied to another existing ground. Shouldn't it be able to exist on its own and provide an independant 0V reference?
Also why do you suggest going to the PSU first and then the star ground from there? Why is that better than both points connecting at star ground? Just curious.

 

[bcarso]

Look at the circuit and see when and where the big currents flow. When rectifier diodes charge capacitors there are big current pulses. These create voltage drops in wires/traces.

A star ground in the real world has its own local voltage drops based on the distributed resistances--look at all possible paths for the current flow. If you confine as best as possible the charging current blips to the vicinity of the caps and then go from there to what you are declaring to be your star ground, you minimize the bumping of things tied to different parts of that junction by the cap/diode junction.

In amps I've done it may only take a small fraction of an inch of trace from the cap common to the star junction to reduce the hum/buzz by many dB. It is hard to get layout people to do this for you however, particularly the ones who believe they understand how it is supposed to be.