Faux ground locations for 6.3v heaters

[Helmholtz]

There is indeed a tiny heater ground current.
The main source of this current is the HT winding coupling to the heater winding via inter-winding capacitance.
The capacitive coupling favors the higher harmonics of the distorted HT voltage.
(Capacitance values vary from 100pF to over 1nF.)

The current loop is HT winding > capacitance > heater winding > tubes heater-cathode leakage > ground wiring > grounded side of the HT winding.
Grounding at the reservoir cap means the shortest possible current path and prevents the current from reaching the tubes and/or passing through the ground wiring.

[R.G.]

I wish we could stop calling it grounding. What we’re doing is referencing the heater circuit to chassis potential. The chassis is grounded. The heater circuit has no current that is entered the chassis. We’re referencing it to this point so that it remains set here and isn’t floating. The other circuits with the amplifier may use the chassis metal as a conductor and return current to their source but none of the current in the chassis is moving through the safety earth ground connection that leaves the chassis unless there is a fault.

I wish it would be called something else too; there is a world of mis-perception in calling everything grounding.

I did a lot of looking at "grounding" recently. This left me with some opinions on the topic.

First, there can only be one ground. Ground is a point, not a wire, or a plane, or a shell. Ground is the place you decide will be truly zero volts and all other measurements will be taken referred to this point. By definition, the place you pick for this One True Ground (OTG) is the quietest place in the circuit. Other pl aces will measure the same zero volts only if there is truly zero current flowing between OTG and them. The voltage may be inconsequentially or imperceptibly small, but failing superconductors, there is a voltage drop when current flows in any conductor.

The chassis cannot be ground, it can only be ground referenced. And it can only be ground referenced at all points on it if no current flows through it. The voltage may be small, but all real conductors obey V = I* R.

I think we're referencing heater windings to signal ground, not chassis. In an ideal world, each tube would have its own heater winding and be tied to the local signal ground for the tube it serves. We pretty much can't have a winding per tube, so we fall back on tying the heater CT to someplace quiet and hope this doesn't have too much wobble with respect to the tube's signal ground. Mostly this works if we pick a quiet place to reference the heater. As Helmholz points out, there is always some leakage to/from AC windings, so it's a game of infinitesimals whether the chosen reference point can eat this leakage without wobbling the effect of the heater on the tube around.

Using the chassis as a DC return ground for the circuits in the amp guarantees that some points on the chassis will be at different voltages than others by V= I*R. Whether this causes audible noise depends on the micro-minutae of exactly where on the chassis current is injected and removed, and what signal reference points tap off these voltages exactly where.

A good way to avoid hunting for quiet spots on the chassis to to make the chassis into an RF shield/safety shield with no deliberate currents on it at all. Speaking of RF, RF directly induces voltages across the chassis metal. It's unusual for it to be big, but the chassis itself can in some degenerate situations induce RF signals between circuits "grounded" at different places.

This is just my opinion, of course, but in my mind, I would reference the heater windings to the quietest place - by definition - in the amp, the negative of the first filter cap. That's the place where everything (that doesn't radiate out, at least) comes back to resolve. Signal ground, DC power return, "AC sewer ground", etc. all get returned here. This is not the only way to get an amp quiet enough, of course, but it's at least predictable.

[Helmholtz]

My post above was missing the conclusion.
I added an essential last sentence.

[martin manning]

Bravo!

[trobbins]

I agree with Helmholtz's shortest loop description: a humdinger (pot or divider) will aggregate all the parasitic current from the valve heaters on that heater circuit; and such parasitic current most likely relates to the capacitance between HT winding and heater winding being relatively high, and the likelihood of the forcing voltage waveform between HT winding and heater winding being relatively high.

The general topic of the HT winding inducing hash into the heater voltage waveform, and then using a humdinger pot to suppress the total hum+hash, was the topic of a 2012 diyaudio thread (https://www.diyaudio.com/community/thre ... ly.207017/). In that thread I had a link to an AX84 thread where Merlin had posted a relevant scope plot, but sadly I don't think the wayback machine has saved the relevant thread circa 2008 titled 'Humdinger w/ this relay setup?' and with a thread identifier of 411564 or 426717 (I can see the thread listing but can't access the thread content). That all goes to being aware of how to minimise and constrain ht rectifier transients, to avoid possible noise transfer over to the heater(s) of sensitive input stages, which may then (depending on circuit) ingress into the audio signal.

My view is, assuming the input stage is most prone to such heater noise coupling into its cathode circuit, then tuning the humdinger wiper to the first stage 0V node, compared to a node that may have more noise on it relative to the first stage 0V node, is conceptually better, as it may not just be noise current in the cathode circuit at play, but also heater noise voltage driving unbalanced capacitances to the grid. Yes any humdinger wiper related noise current would flow back to the HT winding and likely be dominated by a path via a first filter cap and conducting diode, but I don't see that exacerbating the noise influence in the input stage.

I made one effort to try and measure the HT rectifier induced noise in a test setup, and identify the influence of 1N4007 and UF4007, and the frequency spectrum signature was down in the weeds for a modern soundcard, so trying to measure influences such as this topic of humdinger connection could well be below any practical measurement scenario of a preamp or amp.

[trobbins]

With respect to HT rectifier related noise transferring over to the input tube, there is a good thread over on music-electronics-forum :
https://music-electronics-forum.com/for ... -in-output

Unfortunately the thread did not identify or progress to using a tuned snubber to see how well that could suppress the remnant diode turn-off related glitch.