Hum and Noise Preventive Wiring

[R.G.]

Yeah, it's been done to death in other posts, including the sticky post at the top, but here's another cut on the topic.

I ran across a document from 2008 where I tried to condense low-hum, low-noise wiring, with diagrams, into one page. Given the recent posts on hum problems, I thought I'd re-read it and maybe update it. Good choice - it took four pages, not one, to do a sensible job of just skimming the topic. Comments welcome. I reserve my right to be wrong.

amp star grounding.pdf

[martin manning]

Nice! A couple of points:

in 7. High Voltage Rectification: The CT wire must connect ONLY to the
negative terminal of the first filter cap...

This reservoir cap negative must also have a ground connection, as shown in the global grounding diagram. In other words, add the star ground to that diagram.

What about the bias supply filters? I would ground them at the star ground since they typically return to the PT CT. I see that in the global diagram, but not in the small one.

[R.G.]

Good catches!! I'll update the pictures to clarify those. Thanks!

[trobbins]

RG, some comments for thought:

Regarding the speaker circuit, it may be wiser to suggest that the speaker circuit is grounded to chassis rather than left floating, as a floating voltage source may be considered more hazardous than an earthed source for a variety of fault scenarios. Also if the amp uses feedback then it may be better to ground the speaker circuit at the feedback circuit local ground point, otherwise the feedback signal may be prone to pollution.

Regarding output stage cathodes, the concern is to locally loop any large audio signals circulating in the output stage (rather than referring to them as pulses) through that first filter cap. I'd also suggest that is the better star point for any bias circuitry, as otherwise the effective bias voltage seen by the output stage may be prone to pollution.

The diagram of the HV rectification seems to show the CT wire as screened or a radiation emitter, or ? It may be worth mentioning that due to the high ac voltage on, and pulse currents in the diode arm wiring, that wiring may benefit from running as twisted pair from the transformer to diodes and to first filter cap, and include the CT wire in that twist, and keep that wiring as far from sensitive circuitry as practical.

Ciao, Tim

[R.G.]

Regarding the speaker circuit, it may be wiser to suggest that the speaker circuit is grounded to chassis rather than left floating, as a floating voltage source may be considered more hazardous than an earthed source for a variety of fault scenarios.

Good point. Actually, taking it through a resistance to star ground is a safer bet for low noise than chassis. It probably should not float, but for reasons unrelated to keeping the insides of the amp hum-free. Or were you thinking of faults outside the amp?

Also if the amp uses feedback then it may be better to ground the speaker circuit at the feedback circuit local ground point, otherwise the feedback signal may be prone to pollution.

I've thought over exactly this point a few times. In the case of feedback from the secondary, The real point to consider relative to ground is not the speaker, jack, it's the transformer common. The speaker is presumably off somewhere on the end of a speaker cable, and we can't help it out much inside the amp chassis. But we really want to have the feedback point know what signal is being sent out to the speaker.
The idea of making the signal side of the output winding and the common of the OT be a differential signal referenced only back at feedback point is analogous to the issues with the input jack. It's different only in being a lower impedance and shorter cabling, really. For an amp that is not otherwise star grounded, especially with random chassis grounds, yes, taking the feedback circuit ground back to the feedback point makes some sense, as it keeps intervening grounding signals off the ground wire. In a star grounded amp, there aren't any intervening signals to create pollutants to inject into the feedback reference ground.
What kinds of pollution were you thinking of?

Regarding output stage cathodes, the concern is to locally loop any large audio signals circulating in the output stage (rather than referring to them as pulses) through that first filter cap.

We may be at crossed terminology. The output tubes alternate phases. If they are not biased fully into class A, there is some portion of their swing (at least on maximum signal levels) that supplies current through only one of them. The two cathodes need connected locally to ensure that the two cathode currents combine to as nearly a replica of the audio signal as possible right at the cathodes. This keeps any crossover current discontinuity off the cathode return wire. In tube amps, this isn't quite the problem it is in solid state amps, but the issue is the same: don't generate non-signal with output devices switching. But I get where using "pulses" may have misled you. And the intent is, as you say, to force the cathode currents to flow back to the negative terminal of the first filter cap.

I'd also suggest that is the better star point for any bias circuitry, as otherwise the effective bias voltage seen by the output stage may be prone to pollution.

Can you describe what kind of pollution you're thinking of? Is this in the general case of the rest of the amp being grounded in a non-star fashion? If so, yes, picking a return for bias that doesn't share ground offsets with other parts of the circuit is a good idea. But the point of star grounding in general is to isolate current returns so that no sections of the circuit have signals injected by ground offsets.
The PT may have made the question moot. If the PT has only a bias tap, then one end of the winding that generates bias is necessarily tied to the CT inside the PT, so you're getting the effect of rectification pulses on the CT wire whether you like it or not, and your only hope is to filter that out external to the bias "winding" and rectification. Isolated bias windings can get you away from this, but they're rare. The advantage that bias circuits have is that they're quite low current, so filtering can be effective.
I deliberately left out some fancier techniques in the name of simplicity for the reader that just wants to know how to get their amp to be quiet. Probably a better setup is with two bias filter caps: one right after the bias rectifier, grounded to the first filter cap negative, and a resistor to a second cap to star ground. The first does the bulk power filtering, the second ties the bias voltage to a quieter place at the system star ground. A deeper analysis gets into whether you want the return currents from the output tube cathodes wobbling the cathodes of the output tubes around when tied to a stable bias voltage, and how much of either effect.

The diagram of the HV rectification seems to show the CT wire as screened or a radiation emitter, or ?

Yeah, that's confusing. No, the intent was not that it be screened. It's just got big pulses on it. I'll see if I can rework the drawing. Good catch. That's what I get trying to reuse drawings from 2008.

It may be worth mentioning that due to the high ac voltage on, and pulse currents in the diode arm wiring, that wiring may benefit from running as twisted pair from the transformer to diodes and to first filter cap, and include the CT wire in that twist, and keep that wiring as far from sensitive circuitry as practical.

Yep, probably is. It really does have to be a twisted trio, including the CT, to work well with the pulses alternating from one half-primary to the other.
Dang it, I really did want to keep it down to one page; then I gave up and went to two, then three, then four. I think that the drawings and text to describe that well will push me over to five pages. Well, I guess if low noise wiring was simple, there would not be so many questions.

[bepone]

speaker ground cannot be on chassis if nfb is used.. speaker gnd is belonging to electrolytic capacitor which is supplying the nfb stage, PI usually, in PP amp. it is a part of nfb signal and need to travel together with nfb wire. grounding on chassis is the worst option.

[bepone]

also why is in provided star topology brought to the star gnd point , node with highest currents? minus pole of first capacitor? better would be to gnd to the star second electrolytic minus terminal, so we can keep isolated charging currents between transformer and first elco from the rest of the circuit.

[trobbins]

RG, thanks for discussing.

Wrt my floating speaker comment, I was just relating that to safety, not to noise/hum. Safety is another issue, and not per se the aim of your description.

Wrt feedback 0V going to f/b stage 0V rather than main star, I'm more thinking of the practical layouts that some may implement that may not be ideal (depending on where the main star is relative to the output transformer and speaker terminals, and other controls that are sometimes inserted such as presence) and that the star connection to the f/b stage may include f/b stage signal current (depending on the local stage supply bypass connection). There is also a chance of ill-defined parasitic signals coupling from the output transformer given it is large and grounded to the chassis with stray capacitive coupling from its primary and secondary side windings back through the chassis link and to primary and secondary side circuits.

Wrt a fixed bias supply, then yes there are a multitude of circuit options seen, and some of them may have common paths, such as a HT CT with its common point at the first filter cap. For that scenario, the use of a second RC filter stage as you suggest may be an earth loop, as the bias supply goes back to the output stage 0V again.

Wrt HT rectification - yes a twisted trio is what I try and aim for as the most practical means of constraining loop areas and external voltage fields.

Ciao, Tim

[bepone]

actually to do the best training on grounding possible, and knowledge achieved is lasting forever, one must use wooden or plastic chassis.for laughing but is true. then when in that chassis amp works without humm and noise, can be transferred to al/steel chassis. so begginners need to think about non conductive chassis, this is level 0 to pass. after all is easy!

[R.G.]

Thanks for the thought-provoking comments. I edited the doc based on Martin's comments, and now I'm tinkering with it again based on TRobbins' and bepone's. I'll post an update at some point.

I found some notes in the directory from 2008 that resulted in the original. The original idea was to limit the discussion to what happens inside a single amp, and to illustrate the results of the thinking that leads to star grounding as a concept. There are other issues in setups other than single-guitar, single-amp, and other illustrations of how to think about grounding and shielding for low noise.

I have two entire textbooks on the topics of low-noise design that I keep referring to from time to time, and I was hoping to give beginners and mid-beginners a quick illustration. These folks are not going to study a textbook; they get their information from scouring the internet. The internet is a great source of random facts and opinions, but not so good for disciplined thinking, although there are some nuggets in there, like raisins in a pudding. This makes me think I ought to produce a companion - and larger! - document to discuss the underlying fundamentals and more complex points, and add a pointer from the base document into the fancier stuff.

I think I also need to lean more heavily on the up-front advice and warning: there are many ways to get quiet wiring. Star grounding stands out because it's predictably low noise, not because it can't be done any other way.

[pdf64]

The thing that bugs me most about grounding is the term, ‘grounding’
I acknowledge there’s long and venerable precedent for it, but I think that the conflation of a circuit’s 0V common with ‘ground’, and hence treating the chassis metalwork as a big zero ohm sink, may tend to cause a wrong headed approach to the topic.

[martin manning]

I think I also need to lean more heavily on the up-front advice and warning: there are many ways to get quiet wiring. Star grounding stands out because it's predictably low noise, not because it can't be done any other way.

Indeed. No need for builders to get lost in a no man's land of "theoretically best" and "more than adequate" (for the intended use). Fender, who was a thoroughly practical engineer, used a production-friendly system of many chassis ground points, but since the basic principle of keeping the high-current and noisy returns away from the sensitive ones was adhered to, it worked quite well, and still does. This is not to say that there isn't room for improvement: By eliminating chassis ground connections of dubious reliability (through pot cases and the brass insert, e.g.), and grouping the returns to form a relatively few paths from "local stars," the contribution of grounding noise to the overall noise floor can be reduced to the point where time and money are better spent elsewhere.