I'm a grounding convert.

[pinkphiloyd]

I just finished a scratch "junk box" 5F4-ish build. I used the same grounding system I've always used without issue* but this sucker had just enough of a 120 Hz hum to be annoying. You couldn't hear it at all when playing but it was audible at idle and it just bugged me. I knew I wouldn't really get a good nights rest until it was gone. I poked around with my trusty wooden shish kabob stick, checked all my joints, increased the filtering, all for naught. Finally moved all the grounds from V1 through the PI to the input jack and left everything else over by the PT. Dead quiet now.

Also, am I the only person that likes to wire their heaters near the end of a build? I always read stuff telling people to wire their heaters first, but I find it so much easier to get to the other pins and then get nice tight heater twists right down to the socket if I do my heaters last.

*The "ground everything at the filter cap that powers it then ground all the filter caps at one point near the PT method."

[Stevem]

Yup, a split ground like that will always be the way to go as if I had to guess I have applyed it to a easy 180 amps over the years that I have worked on and it killed the 120 hz hum to near nothing, unfortunally you can not mod /apply it to every amp with ease!
On builds I do the preamp heater wiring last unless the outputs are 9 pin types which then I do it all last!

[V2]

How many connections to chassis are you using with that 'split ground'?

[xtian]

Which raises the question about the number of chassis connections. Most modern sources suggest one or two points are best (not including the safety earth). But I'm building my first D-style (D'Lite) and there are approximately 503 ground connections to the chassis!

[pinkphiloyd]

How many connections to chassis are you using with that 'split ground'?

Two. One at the input jack and another at the PT.

You should take this post as what it is; an anecdotal observation rather than some kind of evidence, but it certainly seemed to cure my noise issue.

[rfgordon]

I have to agree with the two-point grounding scheme. I use 16 awg (from salvaged ROMEX) as a preamp buss bar. I've also taken to putting the preamp filters on the board next to the appropriate stages, like Valve Wizard recommends. My preamp buss is grounded on the chassis floor with a bolt, and I now use Cliff jacks.

The PS, PT center tap, and power tube cathodes are grounded together.

The only exception to this is that my reverb transformer is fed from the screen filter, but the verb driver cathode R/C is grounded on the preamp buss.

This last amp I built has cathode biased EL84s, and the filament secondary has no CT. I use a 100 Ohm pot with the center lug tied to the RK for the power tubes. That amp is so quiet that you hardly know it's on unless you crank the gain/master enough to hear the tube hiss (which is minimal, since it's a Fender-style reverb amp).

[R.G.]

The bottom line on all grounding schemes is to know what the currents are that are going to ground, and what conductors they flow through.

Two- to 503-connection grounding schemes are only as good as how well they isolate ground-conductor voltages from amplifying stages.

A huge, and often-missed reducer of 120Hz hum is to ensure that the negative wire from the rectifiers (either the bottom of a bridge or the CT of a FWCT setup) goes only to the first rectifier cap. If it goes anywhere else, including maybe especially the chassis, you get a hum that's almost impossible to remove.

[brewdude]

A huge, and often-missed reducer of 120Hz hum is to ensure that the negative wire from the rectifiers (either the bottom of a bridge or the CT of a FWCT setup) goes only to the first rectifier cap. If it goes anywhere else, including maybe especially the chassis, you get a hum that's almost impossible to remove.

Are you saying that the center tap/negative wire should be connected to the reservoir cap but do NOT connect it to the chassis?

[Phil_S]

I wouldn't want to speak for someone else, but I think the accepted wisdom is the high potential grounds include: 1st B+ cap, PT CT, power tube cathodes, output jack/speaker, and something else I'm sure I forgot. I normally gather them together and ground at a single lug on the chassis. I doubt anyone would suggest leaving it floating.

[R.G.]

It gets back to two fundamental grounding concepts:
(1) All wires are really low-value resistors
(2) All wires make a voltage across them in keeping with Ohm's law.

Whether that voltage ( = current times wire resistance ) causes you problems depends on the product of current and resistance, how big they are) and where the voltage appears for your circuit to pick it up.

As a (bad) example, let's assume that the first filter cap negative terminal is attached to the chassis, and that the CT of a FWCT power supply is connected to the ground bushing of the input jack. The rectifiers feed the first filter cap with a charging current that's 10-20 times the DC average current being pulled out of it. That current must travel from the CT wire into the ground terminal of the input jack, through any imperfect resistance of the jack sleeve connecting to the chassis, through the chassis resistance, and into the first filter cap negative terminal.

This connection hums like crazy, even if the total resistance of the chassis is small. That's because the current times resistance voltage of the rectifier return all props up the input signal ground pin. The input tube amplifies the ground I*R noise with the signal, because it can't tell the difference.

If you move one wire - the CT, of course - back to the negative terminal of the first filter cap, the rectifier pulses no longer go through the chassis. So the voltage they contribute to the input jack signal ground being lifted/wobbled around by the rectifier currents drops to zero, because however low the chassis resistance is, zero current noise through it is even lower.

It's all about what currents flow in what conductor.

[dorrisant]

R.G., are you saying that your grounding scheme would have the rectifier ground and the first filter cap (reservoir) tied to the chassis at a point by themselves and the rest of the grounds (subsequent filter caps, reverb drive/recovery, preamp grounds, input jacks, etc.) could be tied to a separate chassis connection or split to multiple connections if necessary? Just trying to clarify the above answer.

[R.G.]

I wouldn't want to speak for someone else, but I think the accepted wisdom is the high potential grounds include: 1st B+ cap, PT CT, power tube cathodes, output jack/speaker, and something else I'm sure I forgot. I normally gather them together and ground at a single lug on the chassis. I doubt anyone would suggest leaving it floating.

I got all off on principles in the previous post. Here's what I think should happen here.

The negative terminal of the first filter cap is and must be the ground reference for all power supplies in the amp. The PT CT is the source of big pulses of current that will cause 120Hz hum unless the CT is connected to the first filter cap negative, no where else.

The power tube cathodes have significant current, but not as large as the rectifiers. If these cause a ground shift for an earlier amplifying stage, you get feedback and possibly oscillation. Returning them to the first filter cap negative keeps this from being a problem as their ground offset voltages only affect the power tubes, not something else.

The output jack is driven from an otherwise floating transformer winding. All the current going out of the winding has to end up going back to the other end of the winding, somehow. So to a first order, one end (or tap) of the OT secondary goes to the output jack, the jack return must eventually go back to the other end of the secondary winding, so a wire directly connecting them makes a great deal of sense.

If the output winding voltages were not needed elsewhere, that would be that, and in no-feedback amplifiers, you can attach either side or neither one of the output winding to ground, makes no real difference at audio frequencies. But most amps use voltage feedback from the speaker output to the PI. That means they must share a common ground reference voltage.

The theoretically correct way to do this is to insulate the speaker output jack from chassis, and take speaker return back to the OT winding. You can now connect either the speaker jack sleeve or the OT winding end back to the PI signal/power ground. This wire carries only the current that feeds INTO the feedback point on the PI, which is quite small.

Because this current is small, people get away with simply tying the speaker output jack to chassis, and counting on the path from the PI power ground back to the first filter cap and then on to a chassis ground connection to work. Usually it does, but sometimes you don't get away with it, generally in a very high gain amp. It's definitely something you get a way with, not the right way, though.

The chassis needs connected to signal ground and safety ground somewhere. It's best not to count on either the chassis for low resistance, or the chassis for good, solid, tight, low resistance connections. In an ideal world, there would be exactly 1.00000000000... connections to the chassis. This could be either from a single ground point, like the negative terminal of the first filter cap, or a signal reference point. Generally people use uninsulated input jacks, and this is by default the signal reference point for the chassis.

What the chassis is REALLY needed for (other than just keeping all the pieces from falling apart!) is RF shielding. This works best if the chassis is grounded not just to signal ground, but also to AC power safety ground. This works well if you insulate the input jack(s) and tie them for RF purposes to the chassis with a 0.01uF ceramic cap and tie the chassis to the first filter cap negative, or tie the chassis to the input jack(s) sleeve and nowhere else. The chassis must be RF-grounded somehow.

Except for safety ground. The chassis and all user-touchable metal parts must connect to AC safety ground with a low-resistance, unlikely-to-loosen-up connection that can stand up to 25A of line-frequency current. The normal thing is to take the incoming AC safety ground to the chassis as close as practical to the AC power cord inlet, on a bolt/stud used for nothing else, per standards.

In all of these considerations, the important point is where is the current in the ground flowing from and to, and remembering that the only way not to get a voltage caused by current flowing in a wire is to make the current zero.

[R.G.]

R.G., are you saying that your grounding scheme would have the rectifier ground and the first filter cap (reservoir) tied to the chassis at a point by themselves and the rest of the grounds (subsequent filter caps, reverb drive/recovery, preamp grounds, input jacks, etc.) could be tied to a separate chassis connection or split to multiple connections if necessary? Just trying to clarify the above answer.

I would and have done this:
The rectifier negative wire connects to the first filter cap negative.

A single wire goes from this point to the chassis. No other chassis ground connections except input jacks and possibly the speaker return, although I don't like to do that.

The first filter cap negative point is the master ground. A circuit "cluster" ( a tube stage or two which work on the same signal and from the same supply voltage, plus the decoupling cap serving them) is wired into a local ground clot, and that clot is referenced to the master ground with a wire.

In particular, the power tube cathodes are collected out by the sockets, and a wire run back to the master ground point. That clot uses a lot of current, so the drop across its ground wire is bigger than the other clots.

A preamp channel, being one or two tubes, is a clot, and has a local gathering of grounds, and a power/reference wire run back to the master ground.

The places to really watch are the high gain inputs: channel inputs, and reverb return.

A group of tube sections working all on the same signal alternates signal phase, being true one section, inverted the next, so the ground currents tend to cancel, if imperfectly. And if they don't, there's rarely enough gain or phase shift to make this matter. It just changes the audio gain fractionally.

So no, there's no remote chassis collection of grounds. Only one master ground point, with related sections branching out of the one.

You'll notice that I've studiously avoided saying the word "star", as there is a cadre of people that will instantly say "star doesn't work", "buss is better", and other things that are even less complementary.

This scheme is a star of sub-stars, which minimizes the sheer number of wires from a pure star ground, but offers nearly as good performance, and almost as much predictability.

I've got some diagrams somewhere.

It is quite true that other grounding schemes can achieve as good performance as a star ground setup. However, there is no way short of modelling or building them to know ahead of time. Star is predictable - if exhausting.

And it goes back to the maxims: know what currents flow, and make them flow in conductors that cause no harm.

[dorrisant]

R.G., thanks for the detailed explanation. I wouldn't be the only one that would appreciate drawings.

I have used the star ground in various forms... also used buss grounding many times too. I have never tried both ways in a similar build though. I need to do that... experiment. I have been doing a lot of DC filaments lately to try to cut the noise to a minimum. Just to try something different, the last build was all AC with a buss grounding scheme and it is super quiet.

I have a recent Hoso build that will get some attention in regards to your explanation. It is not bad as far as noise but I would like to see how low I can get it. I need to build one where different schemes can be tested... I'll put that on my list.

It is great to read everyone's theories on grounding 'cause there's more than one way to skin a cat.

[Stevem]

Some of you are failing to mention the output screen and landing its ground with the main filter!
Blackface and latter Fenders are always made landing this in the wrong place for minimum 120 hz ripple!