Input jack grounding

[R.G.]

That's a good way to do it, as long as the grounds are separated, through the resistance of the chassis.

Why is that?

The chassis may be 6, 8, 10 ohms or even more. Depends on the material. that's why you don't need the resistor.

Why is that?

Sometimes they use a diode, in an effort to 1/2 wave rectify the AC on the chassis. This cuts the hum in half...

Why is that?

Sorry, there is not 0 volts on ground, or the chassis, even though theoretically it should be...
There is always some small amount of AC on the chassis, or on your ground conductor. And that AC is amplified.

Why is that?

But it's important to understand this stuff, especially that star grounding the power supply and audio to the same point is NOT the best method.

Why is that?


As you said, it's important to *understand* this stuff. I'd like to learn. Please 'splain this to me.

[raiken]

Sometimes they use a diode, in an effort to 1/2 wave rectify the AC on the chassis. This cuts the hum in half...

The diodes are not there to "half-wave rectify the AC on the chassis", they are the fault-protection component of "loop breakers", which add a relatively high (10 ohm) resistance to the ground path to reduce loop currents. If the 10 ohm resistor burns up or conducts too much current, the diodes will kick in to keep the ground circuit at a safe potential.

For a more thorough explanation, see here: Earthing (Grounding) Your Hi-Fi - Tricks and Techniques , under the section "Use of Loop Breaker Circuits",
or here: https://sites.google.com/site/asa220...rcuitsdesigned ,
or here: Secrets of hum elimination plug « Tomi Engdahl’s ePanorama blog
or here: http://www.google.com/url?sa=t&rct=j...hjX0ZA&cad=rja

[GainFreak]

Hi,

I recently had a TSL100 for retubing and what I noticed was that the loop breaker circuit (the two back to back diodes/resistor/cap combo) was connected to the chassis at the input jack. Usually these are recommended to be connected to the first filter cap ground. Other Marshall schematics show the same thing. After reading this article I got even more confused:

Earthing (Grounding) Your Hi-Fi - Tricks and Techniques

I've read recommendations the mains transformer core to be isolated from the chassis (via plastic shoulder washers) but the article states the opposite.
Also what if you have a toroidal power transformer? How will that affect connecting the loop breaker circuit if at all? I've seen resistor values from 2R2 to 100 Ohms. How do you make that choice?
Would someone care to explain what is the grounding philosophy behind the grounding in the TSL100? I see there all sorts of "to chassis" connections. There's at jack input grounding (page 5) but in the same time you can see a connection to the chassis from preamp's ground (page 1). Wouldn't that bypass the circuit in question?
Finally is grounding the loop breaker circuit to the input jack right or wrong or it's just another grounding technique that works?

TSL100.pdf

[R.G.]

I recently had a TSL100 for retubing and what I noticed was that the loop breaker circuit (the two back to back diodes/resistor/cap combo) was connected to the chassis at the input jack. Usually these are recommended to be connected to the first filter cap ground. Other Marshall schematics show the same thing.[...]

Would someone care to explain what is the grounding philosophy behind the grounding in the TSL100? I see there all sorts of "to chassis" connections. There's at jack input grounding (page 5) but in the same time you can see a connection to the chassis from preamp's ground (page 1). Wouldn't that bypass the circuit in question? [...]

Finally is grounding the loop breaker circuit to the input jack right or wrong or it's just another grounding technique that works?

Well, if it works, how can it be wrong?

Grounding gets all of this mysticism and confusion because it relies on two things that are subtle: the exact path of ground currents and cancellation/nulling.

Ground issues happen in the first place because ground serves multiple requirements. It serves as [at least!] a reference voltage, a shield, and a return for the power supply.

Ground as a reference voltage needs to be the same voltage everywhere in the mass of circuits. By definition, electronic circuits have no "rejection" of noise on their reference voltage. So if two portions of the circuits need a ground reference, and the "ground" that appears to each one is a slightly different voltage, those voltage differences appear as mixed with the signal, and you now have ground noise.

The only way to have truly zero voltage across a conductor is to have the current through the conductor be zero. All conductors (at least the ones we get to play with, non-superconductors) have some resistance and inductance. So any current through them causes a voltage to appear. This is usually not an issue in high current circuits, like the cathode returns from the output tubes, because another few millivolts on the ground side of the output tubes doesn't cause noticeable signal to appear.

Where you get into trouble is when you connect a voltage reference "ground" like an input jack bushing to the "high" side of a current-return conductor. When that happens, the high-current induced ground-conductor voltage is added to the reference ground and amplified by whatever was using the voltage-reference ground. The same thing can happen with shields. The outer metallic casing of signal cords and enclosures serves to intercept incoming RF and magnetic fields and convert them into currents that run around on the surface of the shield. This keeps them out of the signal circuitry inside. But if you ever set up a situation where you can accidentally sense the voltage difference in different places on the shield by using the shield itself as two or more voltage reference grounds, that reference difference is now amplified.

Things get really bad when you deliberately run high currents through the chassis that you're using as a shield and a voltage reference by using the shield to carry heater currents. The same thing happens on the speaker output of a guitar amp. Lots of current flows in a speaker cable, too. If the speaker jack is grounded through the chassis/shield, an offset voltage appears across that part of the shield that the speaker current follows back to the OT secondary. If your input jacks happen to intercept some of this voltage at their ground reference voltage, you can get both hum and oscillation.

The solution to all of this is to (1) know and (2) think about and provide separate conductors for shield currents and power return currents. Ohm's Law tells us how to make the voltage across a conductor be zero: make its current be zero. So those reference voltage "ground" spots need a conductor to the circuit's global ground point that carries zero current- their own separate wire. If each circuit user of ground has its own wire to the circuit's global ground, the ground currents in each wire are never mixed, and no cross-contamination of ground voltages can happen. This is star grounding.

Star grounding is a PITA. There are huge numbers of ground wires. In practice, you can use cancellation to eliminate some wires. In common with other amplifying devices, tubes do not amplify what's on their input nodes, they amplify the *difference* between the input node and one other node; in tubes this is the grid and cathode. The cathode is where all that plate current returns to the power supply, so the ground-return problem is built into every tube section. We get away with this because that ground-return voltage is **exactly** phase related to the input voltage. It is always perfectly set up to oppose the input voltage change, so it acts like a little bit of local negative feedback. It can't cause noise or oscillation, all it can do is change the gain of the tube section a tiny bit. And we can't muck up the connection because it's inside the tube. The tube has internal partial cancellation of ground return voltage.

This insight leads to two things: a simplification of the mass of wires in pure star grounding, and in why the ad hoc grounding in many amps works at all. If you clump together the grid circuit, cathode circuit, and B+ bypass of a single tube stage in a local star ground, all of the currents flowing into and out of that local ground lump are phase related, and so they may change the stage gain, but they don't introduce noise. A single wire back to the global circuit ground can't cause problems because it's only carrying this phase-related ground difference signal. If you gather each of the other active circuits into local star grounds, all tightly phase-related inside the local star, and take each one of those back to the circuit's global star ground, the return currents do not mix and cause noise and cross-contamination problems. This is the "cluster of stars" grounding scheme that most amps with good noise performance use.

But there are amps that do not use a cluster of stars for their ground and some of them get OK ground performance. Those are special cases of cancellation. From the discussion above, you know that there is a voltage across every ground conductor that carries current. In addition, if you had a sensitive instrument you could sense the voltage DROP along a ground conductor, just like the voltage drop across a volume control. So if you connect a voltage reference ground (like the input jack) halfway between a high-current "ground" return (like the speaker jack) and the global circuit ground, that reference ground gets half the ground shift voltage.

If the main amplifying circuit has any phase shift (... um, got any tone controls? series coupling caps? bypass filter caps?) then moving the voltage sense point ( input jack) along the path also changes both gain and phase shift for the fed-back ground offset. And to add to the confusion, attaching the "ground" for volume controls, tone controls, clipping sections (!) or overdriven stages inserts other ground components into the mess along the way back to the main power/reference ground. That's why I sometimes refer to this as a sewer ground. Anything at all may be dumped into it.

But just like in the rest of life, sometimes you get lucky. You can actually tune the amp's noise, hum, and phase response by moving "ground" wires around on the sewer ground. Some places will cause ugly hum from the heater and transformer induced hum, and some spots will actually *cancel* hum. Some spots can make the thing oscillate, and some others can cancel a weak spot in the frequency response, or actually prevent oscillation by modifying a phase/gain response hole.

And if you've stayed with me for this long, I can now say what I think the grounding philosophy behind the grounding in the TSL100 is - it's the idea of combining some star wires and some cancellation to quiet things down and save money by eliminating "excess" ground wires. They thought and tinkered with the grounding in the prototypes until they got to a low enough noise/hum/stability with a minimum of costly hand wiring. In other words, the bean counters got into the act. It's OK as long as no one upsets the balance.

Amp mods anyone? That could not possibly hurt anything. Um, could it??

I've read recommendations the mains transformer core to be isolated from the chassis (via plastic shoulder washers) but the article states the opposite.

Mains transformers project a magnetic field leaking out of the corners and the ends of the laminations. This can and does cause mains-frequency eddy currents and eddy-current voltages to appear on nearby metal. Also, there is capacitance coupling those many-hundred-volt mains signals to the One True Ground. That causes currents to flow between the transformer core and nearby metal. Connecting/disconnecting the core changes things. This matters a great deal if one of your inputs is "grounded" to the chassis and the path from that jack back to the One True Ground includes some of the chassis that's conducting the leakage currents from the transformer. And it will work differently for each different ad-hoc grounding scheme. So yes, the theory predicts that like blind men and elephants, some people will recommend one, some the other, because their experiences gave different results that they can swear by. And they're telling you the absolute truth - as they know it.

Also what if you have a toroidal power transformer?

Well, you can't ground the core on a toroidal transformer, can you? Fortunately, toroids leak less than any other style of transformer, so it's only the capacitive leakage to worry about.

How will that affect connecting the loop breaker circuit if at all?

Different question: given what I typed, what does a loop breaker DO?

It forces the currents that would otherwise have flowed in that wire section to find some other way around. This works GREAT if it either cuts the ground currents down or forces the ground currents to seek another path so that the local voltage reference is not so tainted by ground voltages that the overall interference is cut. It works not at all if it happens to be inserted into the low current (ground referencing) path.

I've seen resistor values from 2R2 to 100 Ohms. How do you make that choice?

Consult an oracle. Oh, wait. That's what you're doing now. The resistor value is chosen so that with the ground currents that exist in that one instance of the amplifier, not "too much" voltage appears across the resistor, while still forcing the currents to go elsewhere.

Hmmm. I just had a mental image appear. Imagine that you have an amp that's pure, true star-grounded. Every single place on the schematic that nominally connects to "ground" has its own wire to the One True Ground. Those wires, as we know, are actually resistors of varying values.

There is a set of unseen resistors between the circuit elements before their ground wires. These are the "air resistors" between the circuit elements. These air gaps could be any resistance at all, from 100 tera-ohm gaps in dry air down to a couple of megohms in really humid air. Or you could put in a real 10M resistor or so between "ground" points back at the circuit. If you vary the resistance of the cross-circuit-node resistance and the resistance of the ground wires, you get variations of the ground-lift resistance.

And since current flows where it's easiest (lowest impedance), varying the resistances of the paths determines where the current will flow, and that determines where the ground-shift voltage appear. In another sense, star ground makes the resistors between the nodes as big as possible and the resistance between nodes and One True Ground as low as possible, to force the current to flow where you want it.

So the simple way is to tinker. Get the twenty or thirty standard values of resistor between 1 ohm and 100, or as few as you want to try, and sub them in, listening for what is "good enough".

It's a lot easier than thinking.

[GainFreak]

Thanks for answering thoroughly the same questions over and over again! Maybe this should go to the sticky section or something.

Well, if it works, how can it be wrong?

Over the years I've tried different grounding techniques and most of them work but sometimes things work although they are not done "in the right way".
So you're basically saying that there's no right, wrong or "standard" grounding rule? As long as the hum is non present the grounding should be considered "right"?
This looks like the economics. The books can't help you make correct predictions for the future but you can find the explanation of what happened AFTER it happened

[R.G.]

Over the years I've tried different grounding techniques and most of them work but sometimes things work although they are not done "in the right way".
So you're basically saying that there's no right, wrong or "standard" grounding rule? As long as the hum is non present the grounding should be considered "right"?

Well there is a standard, theoretically correct answer: star grounding. Star grounding gets to very low (but not zero) ground noise by rigorously isolating noise sources and inputs from one another. It's an arduous task.

And in another sense, that's corect, there is no "right" or standard answer. There is only a countably-finite number of answers that get to acceptably low ground noise and freedom from ground interaction. There are an infinite number of ways to do grounding wrong - which only means that there is an unacceptable level of ground noise and oscillation susceptibility.

What's different about star grounding is that it's predicable: do it well and you get to low ground noise, no matter what the physical setup or circuit schematic is. All of the other configurations are special cases based on the schematic and physical build of the unit. Some of them will use cancellation to get to their low noise, to one degree or another.

Star grounding can't get to truly zero ground noise, as there is always some current running down every ground wire; otherwise, you wouldn't need a wire there. Cancellation schemes can get to zero, at least for hum and internal noise propagation, because you can ( if you're very, very lucky!) make things cancel out exactly. What I don't like about cancellation schemes is that they're usually fragile: move a wire or substitute a different tube and the cancellation may not be as good any more.

But then I'm a technician, not an artist. A technician insists on predictability, on knowing how to start from somewhere and get to the desired goal by knowledge and skill in applying that knowledge. An artist only cares about achieving that one perfect result, whether it is by conscious or unconscious means, or even sheer luck. One of the worst things to say to an artist or author is "That was truly great! Now make another one that good or even better."

This looks like the economics. The books can't help you make correct predictions for the future but you can find the explanation of what happened AFTER it happened

It's different, but in a subtle way. Economics fundamentally cannot make valid predictions, because the results change based on human nature and wants. Grounding obeys a rock-solid set of laws - but subtle, hard to know ones.

[GainFreak]

And in another sense, that's corect, there is no "right" or standard answer. There is only a countably-finite number of answers that get to acceptably low ground noise and freedom from ground interaction.

I have yet to see a "classic" star grounding in a commercial amp. Multiple grounds to the chassis - that is fairly common and I guess it falls into the "cancellation" category.
If we have to be more specific about Marshall I reviewed a dozen schematics of their most complicated PCB amps and I couldn't see a distinct grounding pattern. It looks like they were experimenting or something which considering their reputation is not very likely. Concerning the TSL100 the ground loop breaker circuit has a 2.2 Ohm resistor which can be considered a short which for practical purposes means the input (preamp) ground is coupled to the chassis with the cap. This is the only amp which has it at the input.
I did some quick tests rewiring one of my power amps lifting the ground connection (it's the only point of the amp circuit that was grounded to chassis) from the first filter cap and leaving only the ground loop breaker (with a 47Ohm resistor). It is as quiet as with the filter cap grounding . Changing the resistor value didn't produce any effect on the hum level - it's just not there.

[Malcolm Irving]

... Multiple grounds to the chassis - that is fairly common ...

That's very true, but I think it's generally a bad idea. Although the chassis is a 'good conductor' the current distribution between any source point and sink point spreads far and wide.
(The pattern is similar to the 'lines of force' between the north and south poles of a magnet.) Not a problem if there is only one source and one sink, but as soon as you get multiple sources and sinks the current distributions are sure to overlap and cause mutual voltage drops. Cancellation could come to the rescue and you can experiment by trying various different positions on the chassis to attach grounds.
For me, connecting the local stars together using wire, grounded to the chassis at only one point near the input jack, in a sensible 'daisy chain' order, works very well. Ordinary hook up wire can be used - no need for thick bus wire. With a wire you know exactly where the currents go.

[GainFreak]

For me, connecting the local stars together using wire, grounded to the chassis at only one point near the input jack, in a sensible 'daisy chain' order, works very well. Ordinary hook up wire can be used - no need for thick bus wire. With a wire you know exactly where the currents go.

In the quick experiment I did the shield of the input jack shielded wire that goes to the main PCB was my grounding wire which maybe is not a good idea but it still worked. Actually when you strip it and twist the shield only it looks thick enough.

[Mike Sulzer]

I have yet to see a "classic" star grounding in a commercial amp. Multiple grounds to the chassis - that is fairly common and I guess it falls into the "cancellation" category.

Forget the chassis conceptually until the end. The real goal is to make sure that no significant currents are in series with the low side of the early stages: that is, grid-to-cathode voltage should be only the intended input signal. A star ground does it, great, but it is not necessary; a sequential set of connections from high level to low level works OK. Just make sure that nothing significant from high level gets in series with low level inputs. Then when you have figured out how to do that, you can say "Oh, by the way, the chassis has to be at essentially the same potential as the low side of signals or you will pickup hum. How can I do that without putting any voltage in series with the low level inputs, given that some current must flow to keep the chassis at the right potential?" Well, the answer can be "connect to the chassis from near the low side of the first filter capacitor." The chassis will not be exactly at the potential of the low side of the input signal (nor will it have the same potential everywhere on it), but it does not have to be, just close.

Any additional connections to the chassis should be shields; shielded cable, whatever. Think of them as extensions of the chassis. Can you use a shield to carry signal current? Yes, but you better know damn well what you are doing, so do not do it. Use two conductor shielded cable with one conductor carrying the low side of the signal current and the shield grounded to the chassis in situations where both high and low input signal must be transferred.

The real tragedy of this is that the guitar cable and connectors have just one conductor and the shield so you are forced to have some unwanted current in series with the input. But that just happened; it was not well thought out.

[R.G.]

I have yet to see a "classic" star grounding in a commercial amp.

I did a clustered star on the Workhorse amps. Worked, first time. Very quiet, no discernable hum at all. I also used isolated input jacks and speaker jacks.

Of course, not many people have ever seen a Workhorse.

You'll probably never see a true star ground. Too many ground wires. I did a clustered star on the main PCB, and that was low enough.

[SoulFetish]

I was hoping some of you could shed some light on a question I have. My first amp design was for a directly coupled output stage (an LTP feeding cathode followers driving 2 EL84s in push-pull). I decided to use independent and separate secondary windings and filtering for the +and - voltages to isolate the driver stage to try to keep the cathode followers supplied with as low an impedance supply as was practical. My question is in how, or where to tie in the grounding from the driver stage to the main amplifier. It has to have an connection to earth/chassis and it references the same 0V ground as the main preamp and power amp... so.... Do I keep it as a separate supply altogether and run the chassis connection near the input jack at the same point as the main ground? or do I tie the two "sewer" reservoir grounds to a common point? here is the driver stage/power supply so you can see what I'm talking about:

18W_Firecracker_PI•driver•output.pdf

[R.G.]

My take is that you have two arguably valid choices. Since current from the +/-150V supply returns eventually to that supply, you want to think about where that current flows. In this circuit, it goes through two paths. One is from +150V to -150V, the DC current through the followers. This does not flow on the 0V wire of the ±150V supply, so it can be ignored as far as grounding to the bigger amp is concerned. The other path is grid current into the grids of the power tubes. The grids are pulled up by the cathode followers, and when they conduct, current from the ±150V supply flows into the grid. That current then flows into the cathode, through the cathode resistors and off to the main supply ground. It has to get from there to the center point ground for the ±150V supply.

Ideally, this current should not get into the signal for the other amps, as it's composed of only-positive half cycle pulses of current on signal peaks. A single wire from the ±150V supply to the star ground would do this if the output tubes also have a single-wire ground to the main circuit star.

Another possibility is to move the 0V point for the ±150V supply to the 0V point for the joined cathode circuits at the bottom of the output stage, at the ground end of "TBD". Since the supply is floating and the current pulses it feeds the output tubes are available for collection there, they can be returned to the ±150V supply there and never enter the main chassis ground wiring at all.

[SoulFetish]

My take is that you have two arguably valid choices. Since current from the +/-150V supply returns eventually to that supply, you want to think about where that current flows. In this circuit, it goes through two paths. One is from +150V to -150V, the DC current through the followers. This does not flow on the 0V wire of the ±150V supply, so it can be ignored as far as grounding to the bigger amp is concerned. The other path is grid current into the grids of the power tubes. The grids are pulled up by the cathode followers, and when they conduct, current from the ±150V supply flows into the grid. That current then flows into the cathode, through the cathode resistors and off to the main supply ground. It has to get from there to the center point ground for the ±150V supply.

Ideally, this current should not get into the signal for the other amps, as it's composed of only-positive half cycle pulses of current on signal peaks. A single wire from the ±150V supply to the star ground would do this if the output tubes also have a single-wire ground to the main circuit star.

This is the approach I was thinking about taking. But you bring up a good point which made me realize I was making a rookie mistake. I had been designing the layout as if the 0V for the ±150V supply was the return. But current will exist along the ±150V nodes through the transformer and the centertap is just the voltage point in this case. Makes sense and is fairly basic, but my thinking was wrong. Shouldn't make a huge difference as far as any physical changes, but I'm glad you mentioned it.

Another possibility is to move the 0V point for the ±150V supply to the 0V point for the joined cathode circuits at the bottom of the output stage, at the ground end of "TBD". Since the supply is floating and the current pulses it feeds the output tubes are available for collection there, they can be returned to the ±150V supply there and never enter the main chassis ground wiring at all.

hmm... this is why I've been obsessing about the layout for probably way too long. I don't want to hijack the thread, but the approach is basically this; In order to accommodate the normal guitar amp interface and common chassis/head cab configuration, I'm using a hybrid star scheme. Each stage sharing a common star directly to the supply filter cap feeding that stage(fairly common approach). Each cap, feeds exactly 1.000000 ground wire directly to the preceding(succeeding?) cap in line on to the reservoir supply. However, the layout I drew up has the common resistor which is shared by the output tubes grounded at the reservoir supply. But according to the approach I was going to take, it probably should be grounded to the screen supply. Right? Or should I run three current path ground wires as well as the 0V voltage reference from the ±150V supply directly to the reservoir; from the PI, from the screen supply, from the TBD resistor, and from the 0V point on the separate supply? I suppose I could run a single wire from each filter cap, but... its a pain in the ass and doesn't look as good. I wish I had a better reason but there you have it.

[Malcolm Irving]

I would go for a combination of fig 15.12 and fig 15.13 from the valve wizard article:

http://www.valvewizard.co.uk/Grounding.pdf

The 0v of your ±150v supply would replicate the way the bias supply is grounded in fig. 15.13.