B+ vs. Preamp Filter Caps

[JoeTele]

I'm going to be starting a 5E8A Tweed Twin build. I'd like to follow the Hoffman grounding scheme, which differentiates between pre-amp and B+ filter caps. Can someone tell me which are which?

Also, all of the 5E8A layouts I've seen show filter caps 2, 3, and 4 connected to each other at ground, while there is no wire running between 1 and 2. Is there significance to this?

Thanks!

Joe

[guitarmike2107]

Those caps are connected to the ground plain, both are grounded to the chassis, causing current to flow in the chassis, I would normally run a wire to connect them rather than ground to the chassis/brass plate higildy pigildy

the power valve caps are the one that connect to the power valves, so that's the first three 16uf caps

cheers

[R.G.]

The baffling things about grounding in general is that exactly where the currents flow matters because all wires (and chassis metal, and...) are actually low-value resistors, and the common (mis-)conception is that there is zero voltage between all ground points.

Where big currents flow, the voltages generated across ground wires make a big difference. Here are some highly important examples:
1. The "ground" wire from the PT center tap must go directly to the first filter cap, no where else. This wire conducts big pulses of current from the first filter cap, so the wire resistance makes the voltage along the wire vary. If you connect this somewhere to chassis, the amount of hum you get will vary depending on where it's connected.
2. The output stage conducts heavy (for tubes) currents from output tube conduction on alternate half-cycles from the tube "ground" to the negative terminal of the first filter cap(s). This wire should not be shared by any of the preamp circuitry, or it will appear as a small distortion signal on each section it shares a ground with. Screen filter cap should be tied to the first filter cap negative for the same reason.
3. Succeeding decoupling/filtering caps for preamp sections may be individually grounded to the first filter cap negative, or connected into one or more local star points, and these points connected to first filter cap negative.

It's likely that 2 and 3 are where the grounding scheme you mention get that.

[gui_tarzan]

RG, you said something that got my attention.

"The 'ground' wire from the PT center tap must go directly to the first filter cap, no where else. This wire conducts big pulses of current from the first filter cap, so the wire resistance makes the voltage along the wire vary. If you connect this somewhere to chassis"

I believe every amp I've worked on, save the old Gibsons with floating grounds, has the PT center tap going to a chassis ground along with the first B+ cap's ground. Am I reading what you said wrong?

[R.G.]

I believe every amp I've worked on, save the old Gibsons with floating grounds, has the PT center tap going to a chassis ground along with the first B+ cap's ground. Am I reading what you said wrong?

No, you're not.

It is possible to get away with many things that are not technically correct.

If the amps had the CT and the first filter cap grounded to the chassis at one and only one point, they're probably OK. Something like 10x or more times the DC output current of the first filter cap goes into the (+) terminal every half-cycle peak. And the same current is pulled out of the (-) terminal, and must flow back through the CT lead to the PT.

If those points (CT and first filter - ) are separated along a wire, the wire has a voltage across it of V = I * R. I is large(-ish in tube amps) and R is small, but there is a voltage. If you connect the ground of the preamp section on the PT end of a wire, the preamp "ground" sees a moving voltage equal to the charging current times the resistance in the wire.

This voltage is 120-cycle and a "buzz", being pulses at 2x the power line frequency. It's not generally possible to get rid of unless you move the preamp ground reference back to the first filter cap negative. If you ground the preamps at some remote point on the chassis and the CT and first filter cap are connected together at one point on the chassis, it may be quiet, as little current is flowing through the (unknown) resistance of the chassis, and what is flowing is from the preamp itself, not charging current of the power supply.

So you have to ask yourself - do you feel lucky today?

There are an infinity of ways to ground an amp. Only a few of those can be demonstrated beforehand to be quiet. Others may be quiet, and can be found by trial and error. Others are quiet by accidental cancellation, and remain quiet - until something changes.

As I said, it's all about exactly what conductor the current flows through. That makes it HARD to deal with for voltage-trained people. Like us.

[JoeTele]

Great! Will treat the first 3 caps as the power supply, keep them together and ground at the OT, and run the 4th to the preamp ground buss going from the pots to the input jacks.

Thanks again!

Jo3

[teemuk]

1. The "ground" wire from the PT center tap must go directly to the first filter cap, no where else. This wire conducts big pulses of current from the first filter cap...

This. Currents "returning" to common point from the main filter cap are the same surge currents that are conducted by the rectifier.

Improperly grounded these become a major noise contributor.

2. The output stage conducts heavy (for tubes) currents from output tube conduction on alternate half-cycles from the tube "ground" to the negative terminal of the first filter cap(s). This wire should not be shared by any of the preamp circuitry, or it will appear as a small distortion signal on each section it shares a ground with.

Additionally, some of these high currents are also directed to the loudspeaker. For same reasons loudspeaker's ground return currents should not share conductors with other ground returns.

3. Succeeding decoupling/filtering caps for preamp sections may be individually grounded to the first filter cap negative, or connected into one or more local star points, and these points connected to first filter cap negative.

As a rough rule of thumb, the current draw tends to decrease significantly towards the preamp stages, and the stages drawing less current and operating on more "solid" DC power supply are also less likely to have issues with less-than-ideal "noding" of ground returns. ...That is, as long as you never mix them with ground returns of much higher currents.


Edit:
Oh, Probably does not relate to particular amp of the OP, but all ground returns with quick current surges (digital logics, relays, switches, etc.) should also preferably kept separate from other grounds and make connection to common only at single place.

[dinkotom]

1. The "ground" wire from the PT center tap must go directly to the first filter cap, no where else.
2. The output stage conducts heavy (for tubes) currents from output tube conduction on alternate half-cycles from the tube "ground" to the negative terminal of the first filter cap(s). Screen filter cap should be tied to the first filter cap negative for the same reason.
3. Succeeding decoupling/filtering caps for preamp sections may be individually grounded to the first filter cap negative, or connected into one or more local star points, and these points connected to first filter cap negative.

So this simply means that, in the end,everythíng is basically grounded at the first cap?

[martin manning]

Maybe a better way to say it is that all current leaving the rectifier must eventually return to the PT CT, or to the rectifier negative in the case of a FWB. The Valve Wizard site has an excellent .pdf on grounding available for download. http://www.valvewizard.co.uk/Grounding.html

[R.G.]

So this simply means that, in the end,everythíng is basically grounded at the first cap?

Yes, and no. That particular limit is called "star grounding".

Pure star grounding has every single component that connects to "ground" have its own wire to the star-ground point. Star grounding can be shown to have the least possible interaction between "ground" points in the circuit, because the voltage across the wire/resistors from components to the One True Ground cannot possibly cause a voltage in other sections of the circuit.

Pure star grounding causes massive numbers of ground wires, and is very unwieldy to wire. So practical - and lazy! - humans looked for other ways.

There is another process at work in determining how much voltage appears across ground wire/resistors - cancellation. In a circuit where each stage inverts the signal, the current from one section is in opposition to the current in the next/previous stage (neglecting phase funnies for the moment). So some portion of the ups and downs cancel. At least, they are the same signal, but added in different proportions. So this cannot introduce distortion or crosstalk, only a change in effective gain by partial cancellation or local feedback.

The partial cancellation idea leads to local star grounds. A few stages with the same signal waveform in them can be gathered into a local ground point with no crosstalk, only a minor gain change. Then these local star points can be connected to the One True Ground star point with little or no loss of signal integrity compared to pure star grounding. This cuts down massively on the number of ground wires.

This works beautifully well. I star grounded a tube amp built on a PCB, and it was noticeably quieter than the same circuit built on wires and tagboards.

Because of cancellation, it is possible to have uncountably many of the infinite number of possible grounding schemes be as quiet as pure star grounding. But it is in general not possible to predict ahead of time which of these non-star-grounded setups will be this quiet. Star grounding is a PITA, lots of wires and lots of thinking (which always hurts!) so humans want to find an easier way. They do something else, don't like the result, then tinker with minor ground wire changes until they stumble on something that's less noisy, and think they have found a new universal truth. They then tend to publish this on the internet and call it the "One True Grounding Way" because after all, they found it worked.

As I said, good grounding involves knowing what currents flow in what wires and then making sure that the currents flowing through these resistors (all wires are resistors!) do not induce unwanted voltages in other parts of the circuit, especially and most critically the input of the whole amplifier.

Teemuk brought out the next lecture early: speaker return grounding. He says:

Additionally, some of these high currents are also directed to the loudspeaker. For same reasons loudspeaker's ground return currents should not share conductors with other ground returns.

And this is correct. The highest currents in a tube amp are the heater wiring and the speaker output. The heaters are set up to be push-pull and center tapped, with no center tap ground current, so this is pretty quiet if done properly. The speakers are not.

The speaker output in a tube amp is usually transformer isolated. The speaker output is push-pull, and has no particular need to be connected to ground at all - until you introduce feedback from the speaker output to an earlier part of the circuit. Most amps do some of this, with feedback usually going to the phase inverter.

This forces the speaker output to share a ground with the tube circuits, otherwise there can't be any feedback signal. So one end (or a tap!) of the speaker winding has to be grounded. This is often done without any thought by simply screwing down the output speaker bushing to the chassis, "grounding" the common side.

But look at where the current goes. If the speaker output winding is tied to the output jack, all is (probably) well, because the return current from the speaker jack common side is sucked right into the speaker winding. Things get ugly if the speaker winding common is just attached to the chassis somewhere else. Now the amperes-big speaker currents have to flow through the chassis resistance to the speaker winding return, causing a voltage across the chassis. If the input jack happens to stand in the way somewhere along that path, the input "ground" is now wobbled around by the speaker return current-voltage. You have unintended feedback from speaker output to input, and a good chance of oscillation.

The right way to do this is to ensure that speaker return currents only flow directly to the speaker winding by taking the speaker winding common directly to the output jack on a "ground" wire that is not the chassis and not shared with any other ground. Now the winding currents have to flow back on their own resistor/wire. The speaker jack can then be grounded and the speaker "hot" output can be taken back to the PI for feedback without the speaker return current/voltage being involved. The theoretically correct way to do this is to isolate the speaker output jack from chassis, and run a speaker jack ground wire back to one of the One True Ground or to the stage that is using speaker-voltage feedback.

The chassis is a substantial hunk of metal, and it's natural for humans to decide that it's all ground, so tie everything to it. That works OK sometimes, miserably poorly other times, and the differences lie in what currents flow in the chassis. If the currents are all small, chances are that the chassis resistance is low enough to make only insignificant voltages; or if the currents flow at right angles to one another, the voltages are lower. But with enough gain, as in the metal-head's cry for "MORE GAIN!!" you can make anything oscillate.

The theoretically correct thing to do is to isolate input and output jacks from chassis, and have one and only one wire connect the chassis to ground so it can act as an RF shield. Done this way, the chassis simply cannot be carrying any currents, so current-generated voltages across the chassis don't exist. In practice what happens is that amp builders use the chassis, and then go looking for places on the chassis to connect certain ground points that are quieter somehow, so they look for cancellation quiet places on this two-dimensional mess of voltages on the chassis. It's not that you can't find such cancellation points, but they're very, very hard to predict.

And then there's the difference between power line, audio, and radio frequency signals. Radio frequency signals can be take to be "any frequency where the field effects cause by high frequency overwhelm the pure Ohm's law effects of current flow". At DC and up to above audio, the field effects are negligible, so you want wires for ground currents. At RF, you need transmission lines and that means planes. Field effects make return currents flow independently under the conductor that carries the signal current, so using the chassis as a big, clumsy ground plane is needed for shielding.

That's why if you isolate your input jacks from chassis, you ought to use a good RF capacitor from the input jack ground bushing to the chassis right there at the jack. This connects incoming RF from the input cord/antenna to the chassis/shield and keeps the RF from flowing down the wires to be amplified.