I've seen 3 different recommendations for routing filter grounds to reservoir

[SoulFetish]

In the hybrid/multi star ground scheme which often makes the most sense given the natural physical dimensions in a guitar amplifier, I've seen conflicting advise as to the proper way to route ground connections back to the reservoir main star ground. The are illustrated as follows:



So, is there a consensus as to which one is best?

[soundmasterg]

Of those three choices, I vote for number 3. I have had good luck with that approach in several amps.

Greg

[loudthud]

I use the top diagram using the minus sides of the caps for local stars for the circuits in that stage. Not shown is where you tie chassis ground. I usually tie chassis ground at the input jack but you need to run a large wire big enough to carry any fault currents.

[uneumann]

I use the top diagram using the minus sides of the caps for local stars for the circuits in that stage. Not shown is where you tie chassis ground. I usually tie chassis ground at the input jack but you need to run a large wire big enough to carry any fault currents.

^^^^ Is my choice also

I think the OP intended it, but I'll make it clear - there is no reason all the caps have to be close together as drawn. The connections remain the same even if you distribute the caps in the layout. In fact, there is good argument to do that - but it's another topic and beyond the thread question.

[SoulFetish]

I use the top diagram using the minus sides of the caps for local stars for the circuits in that stage. Not shown is where you tie chassis ground. I usually tie chassis ground at the input jack but you need to run a large wire big enough to carry any fault currents.

Good! I'm glad to hear you and uneumann have had success with this layout.
I just rewired my circuit ground tonight according to this scheme. I also opted to ground near my input jack via 16AWG wire.
*Hey this is off topic, do you guys have a recommendation and source for polystyrene caps?*

[trobbins]

+1

[Gaz]

First one with ground at input jack. Never fails.

[Malcolm Irving]

I think this is a difficult question, because the ground connections have to carry a mix of two currents.

The first current is power supply return current. In this respect, option 3 seems the best, because the ground voltage at the preamp star point is not modulated by any ripple in the power supply return currents from the PI and Screen supplies.

The second current is signal ground return. In this respect option 1 seems the best as the signal ground return currents do not have to ‘travel’ all the way across to the reservoir star point and back. (Although, having said that, there is no signal ground return from the power stage back to the PI, since the signal between those stages is balanced.)

I haven’t any theoretical or experimental evidence either way, but I’ve always used option 1 and have made some amps with negligible hum and buzz.

One interesting experiment would be to build an amp using resistance wire for all the ground wires, so that the levels of buzz and hum, of the various options, would be more audible and easier to measure. But maybe life is too short.

[Malcolm Irving]

...
there is no signal ground return from the power stage back to the PI, since the signal between those stages is balanced.
...

Keeping that in mind, it might be that option 2 could be the best!

[Chuck H]

Individual ground leads to star point. All circuits share ground lead with their own filter. If using a buss then the main filter should be grounded at the far end away from the preamp with subsequent filters approaching the input jack end in tandem.

JMHO

[R.G.]

Malcolm, you're already on the right track by being aware that you have to worry about what currents flow through the caps.

In grounding, there is nearly always a compromise between what is theoretically ideal and what is practical to accomplish. You're poking right at one of those boundaries. The diagrams are incomplete, even assuming you add chassis ground to it. For better decision making here, add to it resistors as lumps representing the circuit sections, and resistors representing the wire resistances for ground return wires.

This helps you see what currents actually flow. Adding the circuit sections as load resistors makes you think about where the actual operational currents go. That clarifies the question, at least to me. The interstage decouplers carry a mixture of power supply ripple current and signal currents, as the stages each pull more/less current amplifying the signal through them. The question then becomes whether the interstage caps should return that mixture of currents local to the circuit section where the bypassed current will mix with the signal amplification current to return to the power supply star ground, or whether it should return to the star ground all by itself, isolated from the signal amplification currents.

In the case of the interstage caps, their primary function is to de-couple one stage from the next, so the power supply currents from one stage are not coupled into the next stage as modulations on the power supply. If more than two stages interact through the power supply, it can be the source of motorboating - phase shift oscillation at the low end of the pass band. An interstage decoupling cap is mostly about shunting signal currents back to the reference ground, as the ripple voltage at stages earlier than the phase inverter is very small.

So I'd probably vote for number three. It pretty much guarantees that any signal currents modulating the power supply lead get shunted right back to the power supply, not passed to the earlier-in-the-signal-chain stages, which happen to be later-in-the-power-supply chain. I'd also vote for each local star to have its own path back to reference ground for avoiding high frequency oscillation from feedback through the ground resistances.

But grounding is perverse. You can also get mid-band cancellation of some of the signal currents in both ground and power because we use inverting stages. Each stage tends to pull max current when the current in the preceeding stage is minimum. That's good for power supply ripple in midband, but the phase shifts at high and low ends make this progressively less true with higher and lower frequencies out to where the phase shifts at bottom and top eliminate the notional cancellation of an inverting stage.

On the >third< hand, the gain has to be right to make oscillation, or funny "resonances" that are not truly oscillation or cancellation happen. I'm convinced that some of the mysterious "voice" of certain amplifiers may be attributed to the in-band resonances that not-quite-oscillation can do.

And if the ground wire resistances are low enough, any of these is fine, although "enough" depends on the amount of gain and how much current gets pulled. Each person's experience counts hugely in what they think is "best". If something has always worked for you, then it's the One True Way. Until it's not. Really, that's the reason we have theory at all. To get to good results every time, you need more than the collection of anecdotal opinions an an accumulation of what used to work. That's good if you do the same thing all the time, not so good if you want to do something new.

[Malcolm Irving]

If we have a stage (A) fed from power supply capacitor A and then the signal is passed via a coupling cap to a subsequent stage (B) fed by supply capacitor B, we would like the local star ground of both those stages to be (as near as possible) at the same voltage. Otherwise, any voltage difference between those two local star grounds will appear in the signal going into stage B. But, the ground connections of cap A and cap B must also carry power supply ripple current, which via the resistance of the ground connection(s) will introduce a buzz voltage into the signal for stage B. We would therefore like the ground wire or wires connecting the local star grounds of stage A and stage B to be as short (and low resistance) as possible. (As in option 1.)

[nickb]

How about this. Given the opposing goals driven by the dual duty of ripple filtering and signal bypass it would seem desirable to eliminate one of those. The power supply ripple could be almost completely removed by using either very good RC filtering up front or more simply with a capacitor multiplier. With the ripple currents more less eliminated you can now choose your grounding scheme by considering only signal currents.

[Chuck H]

But, but... There are some that would argue too much filtering stiffens the power supply and changes the feel/tone in a negative way!

I've gotten beyond this by using enough series resistance in the pi filter to add some "sag". Sounded good to me. And a lot quieter as a bonus.

I always enjoy R.G.'s ground tutorials, but I wonder how many times he's had to cover the same material. If we gathered all his grounding posts and edited for duplicate content it would still be quite a tome, but a terrific resource and much easier for him to link rather than rewrite it five or six times a year. This time he added the aspect of viewing current blocks and ground paths as resistors. Which is how I've always done it and it's never steered me wrong. So that was a new twist I like. Ok then, let him keep writing about it

[trobbins]

When using style #1, I'd just like to reinforce the notion of using a practical but not too small 0V wire (as noted that wire needs to cover any fault current), and to make stage layout such that the 0V wire path is as short and direct as practical, and to make sure to make each local star node as close to the bypass cap internal terminals as possible. That makes style #1 approach style #3, and allows each stage to locally loop their own signal current with their own bypass cap (to avoid parasitic coupling from longer cable runs criss-crossing the amp).

For style #1, the 0V node that is grounded to chassis becomes a dividing line for parasitic ground/earth currents generated internal to the amp (eg. from capacitances between power transformer windings to chassis), to those originating external to the amp (eg. from equipment and cables connected to the signal input socket). Making the chassis connection at one end of the ladder or the other end of the ladder may be worth testing for, but the risk of problems from external currents usually pushes many to ground at the input socket.