grounding multiple choice

[exclamationmark]

Another example that's frequently a problem: the wire from the high voltage winding center tap carries big currents, and carries them in sharp pulses once each time the AC power line peaks. This wire ought to go directly to the negative terminal of the first filter cap, and nowhere else. If the signal grounding of the amplifier shares any portion of the wire going between the HV center tap and the first filter cap negative, then those current pulses show up as a voltage across that section of wire, and they're big enough to jiggle the "ground" for the rest of the amp up and down at twice the power line frequency. So you get a buzzy hum that's impossible to cure unless you figure out or stumble into changing that wiring.

Actually I have a question about this - since you can model the wire to the centre tap as a small value resistor, what stops the rest of the circuit from floating up and down when the current does flow through it (and presumably induces a voltage across it)? If you make the first filter cap the main star point, won't all the other star points that are attached to it bounce around as well? Or does the the fact that it's referenced to the earth via something very low impedance eat up the noise?

[trobbins]

Heater induced noise is mainly associated with the input stage. If I had the option, I'd take the humdinger to the input stage 0V node (however that is done), as it minimises adding any 0V noise that may develop between the distributed 0V points.

[Bruce / Mission Amps]

Actually I have a question about this - since you can model the wire to the centre tap as a small value resistor, what stops the rest of the circuit from floating up and down when the current does flow through it (and presumably induces a voltage across it)? If you make the first filter cap the main star point, won't all the other star points that are attached to it bounce around as well? Or does the the fact that it's referenced to the earth via something very low impedance eat up the noise?

Hmmm... OK, where do the electrons come from or move from with respect to the power supply providing the joules of energy required to make the speaker move, and subsequently , you hear it...?

[R.G.]

Actually I have a question about this - since you can model the wire to the centre tap as a small value resistor, what stops the rest of the circuit from floating up and down when the current does flow through it (and presumably induces a voltage across it)? If you make the first filter cap the main star point, won't all the other star points that are attached to it bounce around as well? Or does the the fact that it's referenced to the earth via something very low impedance eat up the noise?

Good question. It gets back to the issue of - what is ground, really?

"Ground" originally meant something like "connected to the planet itself". We still use this in ground rods for electrical distribution and safety ground.

The modern definition of "ground" is more like "anywhere in a circuit you decide is your reference voltage of 0.0000... volts". You're free to decide that ANY point in a circuit is real, true ground. And, barring interactions with the AC power lines and leakage on the power lines, which is another issue entirely, you can run a thick copper wire from any point in a transformer isolated circuit to a metal rod driven into the ground, and nothing bad will happen. At least until you try to force some other point to be at the same voltage when it doesn't want to be.

The short answer to your question is that if you define the first filter cap negative lead as ground, and tie all your signal grounds to it, then they are at that ground potential and stay at 0V (excepting any current induced voltages from the signal grounds conducting current back to that star point. It's the transformer CT that wiggles up and down with the rectifier pulses, because you have decided to reference everything to the first filter cap negative lead.

That seems simplistic until you realize that there is NOTHING else connected to the transformer CT. It can wiggle itself up and down, and does, but since there is nothing else connected to it, it's playing wiggling games by itself and causes no other signals to move around, so any noise/hum it has on it does not get into the audio signal path. The transformer CT moves relative to the first filter cap negative pin because of the currents in the CT wire. Which one moves depends on which one you define as standing still.

If you say "my first filter cap negative lead is stable ground and does not move", then you connect both the circuit signal ground points to it and the transformer CT, then The circuit grounds follow the first filter cap negative, and the transformer CT moves with respect to the first filter cap negative reference point. All is well.

Where you get hum is when you tie your signal grounds to any point on the wire from the transformer CT to the first filter cap negative. NOW you get the voltage pulses of the current pulses times the wire resistance prying the signal ground away from the power supply ground point on the first filter cap, and that appears on the signal ground as hum.

I'm obviously talking around the issue a lot to get the point across. Any two points on a conductor can only have the same voltage in two cases. The first is if the conductor is a superconductor, and has truly no resistance at all. I don't dunk my amps into liquid helium or liquid nitrogen and have copper-yttrium-lanthanide conducting wires, so my amps don't have superconducting wires. The second way to have zero volts between two points with real, resistive wires, is to have zero current.

If one of those is not true, there will be a voltage difference between the two points of V = I * R.

In the case of the rectifier/filter setup, it's important for low hum to have one wire connecting the CT (or negative side of the rectifier bridge in a full wave bridge setup) go only to the first filter cap, not to the signal ground and not to the chassis. The transformer winding can then wiggle up and down with the rectifier current pulses and it does not drag anything else up and down with it.

I guess to sum up - ground's wherever I say it is, not where the transformer CT is. If I define it right, and isolate currents to conductors so the voltage from currents flowing in the don't disturb my input reference voltages, I get low interference and hum. The trick is to know the currents, and know what they're used for. After that, you can get as quiet as you want to by jiggering what current flows on what wire.

[exclamationmark]

I think I understand it a little better now (sort of..) Hypothetically, what would happen if you chose some point between the centre tap and the main filter cap as the signal ground AND power ground (by power ground I mean this point is referenced to the chassis) Since I've defined it as my 'ground' point, the voltages at the filter cap should swing positive of this point and the C.T should swing negative (by a few millivolts due to the charging current), right? Also, because this is my 'ground' point, no hum would be induced in other points directly connected to this one via charging currents (at least in my reasoning ). Is there any specific disadvantage of doing things this way, or will it just fail abysmally? I ask because this method would be somewhat conducive to PCB routing.

[trobbins]

Main filter cap feeds other supplies, so the bit of wire between your star and the cap neg leg is common to all other supplies - and it has rectifier noise on it - so not good.

[R.G.]

TR's correct. Wires are resistors. Some point midway along the wire(s) between the transformer CT and first filter cap negative will be a voltage divider, and will give you part of the current-induced hum from the rectifiers.

In a tube amp built the way they most commonly are, you're stuck with the negative of the first filter cap being where the signal ground is tied. Ultimately, this is because the amplification stages before the PI are single ended, with their cathodes and biasing tied to the first filter cap negative for DC and operating current reasons. So if your signal ground is tied anywhere else, you will see the effect of any voltage shifts on the wires connecting signal ground to the first filter cap negative. This is an issue I realize I did not elucidate in my first polemics.

It is possible to design the circuit differently, so the signal ground is not inherently tied to the first filter cap negative, but it's just not done that way in the hyper-limited realm of guitar amp design. I've often thought how different (and quieter) guitar amps would be if the early ones had duotriodes available to make the input stage be a diffamp with some common mode rejection instead of being single ended like they are.

If it weren't for this issue, yes, you could probably decide someplace else is "ground" and not have hum added to the audio path. In all cases you're free to decide that anywhere is "ground" for measurement purposes, but this only references the instruments you measure with, it does not change where the currents flow and what the voltages are, it just shifts them relative to the instruments. Well, OK, it can also burn the ground wire on the scope probe if you happen to "ground" the wrong point in a power supply. Ask me how I know that.

In the standard guitar amp, there is one standard answer that works from first principles, although other things may work by hum cancellation or sheer luck in some cases. You tie the CT to the negative terminal of the first filter cap, nowhere else. You then tie one wire from the chassis to one of two points, either (1) the first filter cap negative, having insulated the input jacks from the chassis, or (2) the input jack sleeve and NOT the first filter cap negative. Signal grounds in the mess of circuitry in the amp are then referenced to the filter cap negative, which is where the input jack sleeve ultimately is referenced. If the heaters are run on AC and any point is grounded, ground it with a single wire to the first filter cap negative. The speaker jack should be isolated from the chassis, at least to the point that the return currents don't flow through the chassis to the output transformer secondary winding. You can get away with the speaker jack connecting to the chassis if the OT transformer secondary goes out to the speaker jack sleeve for returns and the speaker currents don't flow through the chassis.

The AC power safety ground is tied to the chassis near where the power comes into the chassis. For safety reasons, this wire is left with more slack than the hot and neutral wires so if the cord is pulled out, the power wires part first. It plays no part in the hum in the amp as long as there is no fault or leakage from the AC power wiring.

If you get the wiring right for rejecting rectifier hum and don't muck up the heater grounding, then you have bought some slack in how the real amplifier circuits treat the chassis. The currents in the actual circuit are small, and if you're lucky they may not add up to make oscillation happen, so you can do something like the brass strip used in some amps under the controls. Especially if you have time to muck around with where what ground wire connects to which on whither when why places on the chassis.

So I guess the short answer is that you might get away with what you propose, but chances are you will induce some rectifier buzz. When one is doing PCB layout, it's easy to slip into thinking that I'll just fudge this to make the PCB layout work. But it's important to remember that PCBs are not there to be laid out easily. They are there to get the wiring right and make the rest of the stuff come out right first time, every time. I do a lot of PCB layout, and so I know the dilemma. Resist the temptation to twist stuff off the PCB to make the PCB easier. The PCB is there to make that other stuff not be twisted.

I did a full-bore star-of-stars grounding PCB for the Workhorse amps, all on the PCB. Essentially, each tube had its local circuits collected together in a local ground, then a trace ran back to the first filter cap negative separate from the others. It came out very quiet. I found that connecting the ground traces remote from the star point induced noise in some cases, not others. It was not easy to lay out, but it did work.

[Gingertube]

For beginners one of the easiest grounding schemes to get your head around, and to implement as you build is the bussed 0V.
You have a 0V buss starting at the input and ending at the power supply main filter cap -ve connection (often also the HV winding centre tap).

You build it with "connections follow schematic",
That is:
1st connection is screen of input cable from the ISOLATED FROM CHASSIS input socket
2nd connection is input tube grid leak resistor ground side,
3rd connection is input tube cathode bias resistor and bypass cap ground side,
4th connection is input stage B+ filter capacitor ground side
etc. right up to the other end which will be
2nd last main filter cap -ve,
last HV Winding Centre Tap.

That 0V buss is then connected to chassis ground at a single point ONLY, usually at the bottom of the input stage grid leak resistor.
If you have done your layout planning well, this will usually be on the opposite corner of the chassis from the mains safety earth to chassis connection or at least well away from it.

ALTHOUGH I have also seen the following places used to do the chassis ground link:
1) The bottom of the grid leak on the Reverb Recovery stage (probably and attempt to minimize noise from the reverb and indicative that perhaps there were other problems which should have been addressed)
2) The bottom of the phase splitter of the power amp, this one was easy to understand, non-isolated (from chassis) speaker output sockets were used and these defined the 0V to chassis connection (because global negative feedback was being used). The the 0V to chassis tie (in this case the 0V is also speaker common) was made at the point where the speaker output was applied as negative feedback, that is, at the phase splitter of the power amp. Note: In this case there was no opportunity to use one of those 10 Ohm + cap + back to bBack diode type links.

There is a bit of story in that last one. 90%+ of ground loop problems are usually associated with Non-Isolated (from Chassis) Input Sockets or Non Isolated form Chassis Speaker Sockets. You can use one or the other if you are careful but using both is usually asking for trouble and using neither is usually best.
Cheers,
Ian

[Steve Conner]

I don't quite agree with RG. In some of my more adventurous builds, I've pulled a "change of reference" between preamp and power amp to help break ground loops. In a push-pull amp, this can easily be done between the PI and the power tubes. The signal here is high amplitude and balanced, so it will reject quite a lot of ground noise. So for example, you can ground the preamp and PI to the input jack, and the power amp somewhere else.

You can do something similar with solid-state power amps. They usually have a differential input stage that will happily float compared to the rest of the circuit. You can separate the input stage and feedback network from the main ground, ground it straight to the input jack and get quite a nice reduction in hum and distortion. I've seen this done in commercial products and tried it myself with good results.

I'm currently trying to work out the grounding scheme for a 24-bit audio player with a hard disk and switching power supply. :-/ I'm really tempted to add balanced outputs using the THAT driver ICs, which will allow me another change of reference.

The usual newbie mistake with star grounding is to send the rectifier current through the main star point. After all, this is what you'll do if you follow the theory to the letter. But the current is high enough that the resistance of the stack of washers and solder tags becomes significant, and rectifier buzz pollutes everything. To fix it, you have to move away from the pure star topology, connecting the rectifier directly to the first filter cap. Been there, got the T-shirt.