Ground layout

[uneumann]

While mulling over a new build, I recall the discussion a bit ago about part layouts. The agreement seemed to fall on the side of "keep high impedance lines short" so layouts should favor part positions near grids or whatever the higher-impedance connection is. Makes sense to me... but what about power and ground layout within a chassis. Many amps (classic Fenders) have controls on the front, tubes along the back, and circuit boards in the middle. The classic ground scheme is to route grounds near the the pots (front) and power towards the tubes (back). That means cathode circuits and grid leaks are relatively far from ground, while plate loads are relatively near power. Does this make sense? In mulling it over, I'm coming out in favor of running grounds near the tubes to minimize path lengths for grid leaks and cathode circuits. Those grounds seem more critical (noise-wise) than the pot connections for tone stack, vol, etc.

History shows that good performance can come from the traditional approach, but perhaps things get even better with ground layouts nearer the tubes.

Any thoughts or arguments either way on this?

[R.G.]

Understanding power and ground wiring requires taking our usual understanding of circuits as being primarily voltage-related, and standing it on its head. To get power and ground wiring clean and non-interfering, you need to think of where the currents are flowing, and what that does to the resistance of the wiring.

The fundamental rule here (at audio frequencies at least) is Ohm's law: current flowing through a resistor generates a voltage. All wires are resistors, and so currents in the power and ground wires generate voltages from one end of the wire to another. The bigger the current, the bigger the voltage.

As an example, take the speaker return wire. This is generally the highest current in a tube amp, perhaps excepting heater wiring. The speaker output comes out of a transformer winding, and must return to that winding. It also has to be referenced to signal ground so the feedback into the power amp will work. If you run the speaker jack return wire back to the transformer, the big currents flow from transformer to transformer, not flowing through other "ground" wires. The ground reference is then taken from the speaker jack or the transformer return terminal, and that wire run directly to the power amp ground near the PI tube (which is usually where output feedback goes). There is only the tiny signal current of the feedback voltages in this wire, so there is nearly no voltage drop, and the feedback is clean.

If you run the speaker return off to some other ground point, especially one shared with the preamp, the current generates a voltage in those wires on its way back to the transformer winding, and if the ground net is such that the ground of a triode gain stage is on the way, you can easily get enough voltage on the ground to make the amp oscillate. Even if it doesn't oscillate, it may have funny resonances in the overall responses where it doesn't quite oscillate.

The output tube cathodes are the next big current drain. The output tubes get their power from the first filter cap, and their cathode currents eventually go back there. If you provide a wire for the combined currents of the output stage to flow back to that first filter cap by themselves, the voltage generated by that current in that wire is not added to a "ground" reference level somewhere else.

If you ground everything to the chassis, the currents make their best way across the metal, interfering and adding complexly as they go.

Using short wires is good, but for ground, which acts as both a reference voltage for the delicate audio signals you're trying to amplify, and a sewer for the return of the "used electricity" from all stages, you need to be picky about *where* the current flows. Short wires are good, but in many cases, a longer wire with only selected "ground" signals tied to it is better.

Don't make your circuits drink from a sewer.

[gui_tarzan]

Sort of puts the "star" configuration at a disadvantage, doesn't it? Tying area-specific grounds together seems like a very logical thing to do, especially with the speaker/triode example. It makes sense.

[R.G.]

Sort of puts the "star" configuration at a disadvantage, doesn't it? Tying area-specific grounds together seems like a very logical thing to do, especially with the speaker/triode example. It makes sense.

Well, it depends.

I am such a tub-thumper for star grounding that I deliberately avoided it in typing that out just to not seem one-noted.

Star grounding is not a panacea for every noise issue, only ground return ones, and it is a royal PITA to actually do in the pure form. In fact, what mostly happens in real-world star applications is that all the grounds for a local clot of stuff that processes one signal are grouped together in a "local star node", where, for instance, all the grounds for the output stage might be collected. That node is then run on its own wire to the global star node.

What that does is to take advantage of the inverting nature of most active devices and add up all the ground currents locally. Sometimes, most times in fact, they will tend to do a sloppy cancellation, where summing them locally produces a smaller AC current back down the ground wire. In the case of the output stage, if you run each power tube back to global star ground separately, each wire contains a half-wave rectified (ish) lump of the signal, with whatever overlaps Class AB causes. If they join at the output tube cathode grounds, they at least come close to adding to a current replica of the signal and having fewer harmonic products to radiate and couple.

However, a pure star on the output stage and output winding is not all that ugly, and isn't a contradiction for star grounding entirely. If we run a wire from the conjoined output tube cathodes to the global star, and also a separate wire for the PI ground to the global star, both output stage and PI think they have several tens of milliohms more cathode resistance. That's OK, as long as the triode gain stages are connected by their own ground wire to the global star. Since each stage has its own cathode wire to ground, the high current stages can't inject voltages by running currents through their ground wires into the input signal ground of the earlier triode stages, and all is well.

If we bring the speaker return ground into this, it's OK if the speaker output winding has its own ground wire to the global star, and so does the speaker jack. That just makes the output winding look like it has a couple of feet more wire between the winding and the speaker jack, and the current goes into and out of the global ground node, leaving a net zero on the global ground.

The feedback to the PI is, however, riding on top of both the output tube cathode return ground voltage and the PI return ground voltage. It's own ground return offset tends to cancel, since the current flowing into the global node is equal in both directions, and if the ground wire resistances are similar, it tends to cancel. In any case, the signal is something like tens of milliohms times amps, or tens of millivolts of offset, and it will be somewhat cancelled. It is also almost exactly the same as the the signal the PI and output stage are already handling, so it's mostly their own ground-line offsets added to or subtracted from the signal that they already have. PI and output stages work on volts, sometimes tens of volts of signal, so the addition of tens of millivolts of feedback, positive or negative, by the ground wires won't have a lot of effect, since the gains from PI to speaker stage are small. Netting it out: there is not enough loop gain from the speaker ground return through the PI ground, output stage, and OT to cause big problems.

That's not the case if you run the input jack ground back to the speaker jack, or to the output tube cathodes. Then the millivolts get amplified a lot.

What this really is is a note that (1) you gotta know what ground currents exist and where they are going and (2) local-circuit ground collection stars are mostly OK as long as (3) the ground offsets on the way back to the global star ground are not shared.

Did I just muddy it up?

[dai h.]

I think you can do a "daisy chain star" (that is, to be more specific), a not strictly impedance separated ground grouping which can work (to my logic) if the currents are low (example: 5F6A/Marshall circuit pot grounds back to input jack as the star point in a daisy chain--super low current returns mean less to get into other returns and be a problem even if the point are not starred). Also the daisy chain(straight line) configuration makes the lowest potential between the points being a straight line(smallest loop area). Also, sometimes on a two-dimensional plane (i.e. a PCB) a star might not appear to be one if it's sideways. R.G.'s local star point is a good/useful one to remember. Another star-ish(to my mind) configuration is a central grouping on a PCB, but the "star" spreading out/pooling out towards the various ground connection ends of the parts (I imagine a thicker foil would help things work better--I understand some of the more "boutique" gtr. amps use heavier copper). I haven't seen a obvious star on a PCB very often but I have one Onkyo (cheap PC speakers) which does have the traces configured in such a manner. I think it's because 1) the currents are high (power amp circuit), and 2) the (presumedly) very thin traces (standard 35um?--guessing) will create a problem if not impedance separated).

IMO, the basics are 1) figure out how basically what you're dealing with works (pickup, tube, transformer), 2) identify the loops (where, how much current, AC/DC(or both), then 3) wire accordingly.

[gui_tarzan]

So then I have to wonder what would happen if you were to put something inline between the different circuits (pre-amp, power amp, etc.) to prevent cross-"contamination" of the ground signals?

[R.G.]

That's not strictly possible if I correctly understand what you mean. Ground connections are inherently parallel, not serial. They must all eventually connect to the One True Ground, and putting anything in series between an outlying "ground" point makes things worse. They may only be trivially worse, as in a few more milliohms in series with a switch contact or thinner wires, but worse.

Putting things in parallel, as in massively more copper wires or much thicker copper wires helps by decreasing the impedance between remote "ground" and the One True Ground.

The special case of running separate ground wire for one section of the circuit to the One True Ground means that you have inserted some milliohms of wire resistance between that circuit section and the One True Ground, but the voltages generated in that wire are kept local to the circuit section being connected. The voltages that ground current generates in that ground wire cannot be passed to any other circuit sections' ground.

... if I understand what you mean, which is always suspect.

[Gingertube]

As usual the "Wiz" has some good reading on this topic with pictures (worth a 1,000 words):
The Valve Wizard
Cheers,
Ian

[tubeswell]

What Ian said