Buss grounding scheme

[Dave H]

Booj posted as I was typing.

[loudthud]

A transformer bolt isn't ideal but why does the order matter LT?

I think he is saying his bus grounding is fine but the original star grounding was no good which was why I was looking for a fault with it.

I've just noticed that the OT secondary isn't grounded on the schematic. I don't know if that would cause hum but it can cause other problems.

Kirchoff's current law states the current entering a junction (the star) must equal the current leaving the junction or algebraically they sum to zero. Link: Kirchhoff's circuit laws - Wikipedia, the free encyclopedia

The idea behind a star ground is that current that enters in one branch and flows out another does not create a small voltage in any of the other branches. There is a small finite resistance between spade lugs in the stack, so there will be a small voltage difference. Depending on the order of the stack, voltages can be induced into sensitive signal grounds where they get amplified. Ideally you should stack the lugs in order from the HV center tap through the resovour cap, output stage, PI to preamp. Unfortunately the nut and bolt would short out from the preamp back to the center tap. Even if you could arrange the grounds to come together at one eyelet, there would still be small voltages on the sensitive grounds.

The daisy chain of stars (as used here) is the best way in practice because it isolates the noisy grounds of the power supply from the sensitive grounds in the preamp. There is a difference in ground voltage between the power amp and the preamp, but that voltage does not enter the signal path in the preamp. Care must be taken in choosing the point to connect to the chassis and the mains safety ground as this can cause a difference in grounds between two pieces of equipment you may wish to interconnect.

[Dave H]

Kirchoff’s laws eh? I think I remember those

Right, I see. That’s a very literal interpretation of ‘star grounding’. Something is telling me that if the stack up order matters then we can’t be doing it right. I’d never connect the PT CT to the star point. It goes to the first filter cap and then a wire goes from there to the star point. I wouldn’t want the capacitor charging pulses anywhere near the signal grounds. I’d do it as R.G. has done it below. Perhaps we need a new name for his grounding scheme to prevent its being confused with literal star grounding which isn’t really a very good way to do it?

star grounding amps.pdf

[booj]

Thanks for [posting R.G.'s drawing, Dave H. I see that my star grounding system was indeed flawed, since it had two ground points, one by the input jack and one by the transformer. (not the transformer bolt). So all I needed to do, to have fixed it would have been to undo the ground point by the input jack. That's some awfully tricky stuff about the order of lugs at a ground point and the small resistances between them. Kind of sounds like a tight buss system. R.G. even has them numbered even though they all go to the same point. concerning the output trans. ground, I tried both ways, running the ground side to the inverter ground node, and no ground at all. The speaker jacks are isolated. Right now they are not grounded. Of course there is no feedback loop either. When I had the virtual filament ground resistors grounded, there was a slight 120 HZ hum. That went away totally when that virtual point was connected to the output tube Cathodes.

[Raybob]

I can't understand why anyone would want to tie the preamp grounds to the star point. I use a buss bar between pots and circuit board with all preamp/input grounds to it, and it's tied to preamp filter caps. It doesn't need to be connected to chassis anywhere, just the grounds of the preamp filters.

[R.G.]

I can't understand why anyone would want to tie the preamp grounds to the star point.

One would do that if one had enough gain and current flow in the preamp sections to cause signal feedback through the ground resistances. This happens more now than it did in the Golden Age because we use more gain in preamps now in some amps.

If we assume the amp has a hum free power supply - perhaps a stack of 24 each 12V car batteries - there is still a grounding issue. This is because the return current in each stage flows through the ground resistors (that, is "ground wires"; all wires are resistors) and causes some small voltage to appear on the ground resistors.

It's easy to set up a situation where the first stage ground has the input ground tied to it, and that is then connected to the second stage ground, that to the third stage ground, and so on for N stages, then to the power supply ground. Put the filter caps wherever in this chain you like.

Ultimately, the power supply filter cap negative is where the electrons come from for all the stages. So the ground resistor/wires come from this cap to stage N. Stage N's ground current flows in this resistor/wire, and causes some DC drop and some AC component, as its current changes with signal. This voltage then is transferred through the ground resistor/wires to stage N-1.

If N-1 has some kind of noise rejection to noise on its ground, all is well. But typical tube stages have *zero* rejection of ground noise, so stage N-1 gets a signal of the sum of all the signal from stage N-2 and the ground noise from stage N. It then amplifies this, and sends it to stage N. Ooops. Signal from stage N is amplified, then sent to stage N's input. That's feedback, and worse yet, unintended feedback. It does unspecified things to the frequency response of stages N-1 and N. If the signal loss in the ground resistors is bigger than the signal gain through stages N-1 and N, we're OK, it just adds some "tone", that being the music industry phrase for anything that sounds different. How big this effect is depends on how big the signal on the ground resistors is and how much gain N-1 and N have.

It gets worse.

We're only at stage N-1. What about N-2, N-3 and so on back to stage 1? Each stage gets a signal on its ground resistors/wires that is the attenuated sum of all the signals from all the stages past it in the chain. The forward gain increases 20-40db per stage for a triode, and there is a lot of phase shift because typical 12AX7 stages only go out to 100kHz or so open loop.

As soon as the multiplication of the gains in the stages equals the attenuation of the signal in the ground resistors, it oscillates at the frequency where the loop gain is 1 and the phase shift makes the ground-resistor signal in phase with the input. Below that, it has odd peaks and nulls in its response from the variation of positive and negative feedback due to the phase of the returned ground signal. This is good if it emphasizes frequencies you like; you can then sell people your "magic-tone" amplifiers which have been imbued with tone-elixir from the touch of your hands. It's bad if it sounds ugly or oscillates.

Note that each stage typically inverts the signal, at least at low frequencies, so you get a mixture of positive and negative feedbacks in some haphazard way which depends on the wire resistances and currents.

There are a few ways to fix this. One is star grounding, which simply avoids the issue entirely by making each stage have a wire back to the power supply. You still get voltages through the ground resistors for each stage, but now these voltages are not mixed and fed back to preceding stages, so there is no feedback. Any hum the stage injects from its heaters stays local to that stage, and is not forwarded back for amplification to the preceding stages. You pay for this goodness by having to run all those #$&*#^ wires and think about how and where the input jacks and chassis get connected to the power supply ground and AC safety ground.

Another way is the bus bar approach. I sometimes think of this as the thousand-pounds-of-copper method. It is impossible to make the ground resistances be zero, but there is some resistance low enough to get the interactions down to minimal. So if you keep pouring copper into the ground path, at some point it gets OK enough. I understand that some people set up buss systems and then hunt for the quietest place on the buss bar to connect each stage. This is literally finding the place on the bar with a cancelling signal to minimize the various grounding uglies. The lower the gain, the better this works. Buss sytems are much easier mentally than having to understand what signal currents flow in what ground wires, and there is some value of pounds of copper that will make it all OK.

About here in every amp design, some amp builder thinks "Hey! The chassis is solid metal! Wouldn't that make a great buss bar? And it's everywhere, so grounding is easy!" and the thinking has come full circle.

Grounding is difficult because you have to think about where the *currents* go and what those results make happen. We all want to think in voltages.

[Dave H]

After reading that I want to make the amp differential right through

[Dave H]

...That's some awfully tricky stuff about the order of lugs at a ground point and the small resistances between them. Kind of sounds like a tight buss system. R.G. even has them numbered even though they all go to the same point.

I could be wrong but I don’t think the numbers represent the stacking order (where would you put the N and the unnumbered wire?). The way it has been done the order shouldn’t matter as the wires where it could matter have been connected directly (like the PT CT going straight to the first filter cap).

[R.G.]

I could be wrong but I don’t think the numbers represent the stacking order (where would you put the N and the unnumbered wire?). The way it has been done the order shouldn’t matter as the wires where it could matter have been connected directly (like the PT CT going straight to the first filter cap).

Nope, you're right. The numbers are just references to keep from drawing wires all over the document. The currents aren't high enough for stacking order to matter. Stacking order does matter in some things, like pulsed-power circuits for rail guns and lasers, but that's a whole different kettle of (fried) fish.

[booj]

The buss system is easier to put in an amp. I wonder if it's really amount of copper or amount of resistance in the wire? Are we talking absolute conductance or distance between points in ohms? Problem no. 1 with the new buss system - my pedal board really is noisy!

[Tom Phillips]

...I wonder if it's really amount of copper or amount of resistance in the wire? Are we talking absolute conductance or distance between points in ohms? ...

The important characteristic is the total resistance between the two points in the circuit that are connected by the wire. There are many factors to be considered if you take things to extremes. However, in general, if a longer larger size wire replaces a sorter smaller size wire AND the resistance of the two wires is the same, THEN the performance of that part of the ground bus will be the same with each wire. I haven't done the calculations to determine if the total amount of copper would be the same in two different size wires of different lengths that have the same overall resistance. The circuit performance depends on the resistance. The circuit doesn't care about the total copper content.

Note that, in other applications, such as signal hookup, the proximity of the wire to other signals is a factor. In those cases, a longer wire could cause a problem depending on how it was routed in relation to the other circuitry and because of the potential to pick up additional noise signal from the surrounding environment. In the presence of high level interference, even ground leads can pick up unwanted signals.

Cheers,
Tom

[R.G.]

Tom is of course correct. I used "pounds of copper" as a bit of hyperbole; the resistance and wire routing is what matters.

Sometimes I just go all literary on you ...

[Russ]

Ultimately, the power supply filter cap negative is where the electrons come from for all the stages. So the ground resistor/wires come from this cap to stage N. Stage N's ground current flows in this resistor/wire, and causes some DC drop and some AC component, as its current changes with signal. This voltage then is transferred through the ground resistor/wires to stage N-1.

If N-1 has some kind of noise rejection to noise on its ground, all is well. But typical tube stages have *zero* rejection of ground noise, so stage N-1 gets a signal of the sum of all the signal from stage N-2 and the ground noise from stage N. It then amplifies this, and sends it to stage N. Ooops. Signal from stage N is amplified, then sent to stage N's input. That's feedback, and worse yet, unintended feedback. It does unspecified things to the frequency response of stages N-1 and N. If the signal loss in the ground resistors is bigger than the signal gain through stages N-1 and N, we're OK, it just adds some "tone", that being the music industry phrase for anything that sounds different. How big this effect is depends on how big the signal on the ground resistors is and how much gain N-1 and N have.

It gets worse.

RG's description of the ground noise problems with N stages etc. got me thinking and I was wondering if connecting one or more capacitors (film type caps) from the bus/signal ground to the chassis ground would bypass this bus/signal ground noise to the chassis ground, thereby removing that noise from the bus/signal ground:



...or maybe it would just create oscillation or who knows what. Don't know, haven't tried it, just an idea. Any thoughts about this?

[R.G.]

RG's description of the ground noise problems with N stages etc. got me thinking and I was wondering if connecting one or more capacitors (film type caps) from the bus/signal ground to the chassis ground would bypass this bus/signal ground noise to the chassis ground, thereby removing that noise from the bus/signal ground:
...or maybe it would just create oscillation or who knows what. Don't know, haven't tried it, just an idea. Any thoughts about this?

Good thinking effort, but it probably won't help much.

Using the chassis for ground returns or signal returns is a problem because the signals get to interact on a big conductor. You can't really bypass the noise to the chassis and pull it out of the bus bar because the chassis just adds more resistive paths back to the ground reference point in parallel with the resistances of the bus bar. Lower resistance is better, of course, but is not a true solution.

In your diagram, instead of drawing the bus ground as a heavy wire, replace each section between dots on that ground bus with a resistor symbol. Also, do not draw the chassis ground as just ground symbols. Pick one and only one of them as the One True Ground, and draw resistors from every other ground symbol to the One True Ground. That's a more accurate schematic.

You can get much the same result by making the bus bar ground out of thicker wire, which is where my "pounds of copper" comment originated. A rod of copper about 2" thick might work pretty well. (As an aside, brass, being an alloy of copper and zinc, has a much higher resistance per unit length divided by area than copper does.)

You also have the input jack attached to chassis ground. That's OK, but if that's how it is done, you want no other attachments to chassis ground except the line/safety ground at the AC inlet. In an ideal world, the chassis would conduct only the currents needed to shield the insides from outside RF and other radiated junk.

[booj]

You also have the input jack attached to chassis ground. That's OK, but if that's how it is done, you want no other attachments to chassis ground except the line/safety ground at the AC inlet. In an ideal world, the chassis would conduct only the currents needed to shield the insides from outside RF and other radiated junk.

VERY SLOWLY this is coming to me. Let me digress a moment.Grounding.pdf
Please take a gander at fig. 13.11. Anybody got any ideas about why you would want to purposely add resistance to the ground side of the power supply?
And Raybob's comment about not needing any ground connection to the chassis at all is very interesting, but I can't help but think there may a be a problem with radio interference.
I see a plethora of things that cause hum, and although my amp is extremely useable, I can still hear it. That deep, 60 cycle MMMMM! lurking somewhere probably in the preamp circuit. It's not loud, and at real world settings the amp is totally quiet, but it's there.
Thank you gentlemen for taking time to explore this subject. And Enzo, the rectifier emulator is the resistor between the B+ supply and the plate supply. Sorry about expecting anyone to be interested in the screen supply resistors. I experimented to find the best sounding values.