Star grounding problem

[Alan0354]

The bottom of the bridge rectifier shouldn't be connected to the power valve cathodes.

The power valve cathodes should be connected to the bottom of the screen supply filter cap.

The SF or indeed BF grounding scheme is nothing to write home about, and is hardly a model example of grounding.

Yes it works, by these are fairly low gain amps.

The grounding scheme seems more to facilitate the construction than optimize the amp for low noise.

You need to force the high current path away from everything else. That is very important and it works. You put large current noise to the screen circuit, you introduce noise to the circuit after the choke. Remember, the ground noise can go to the HV after the choke through the filter cap after the choke. That's a sure good way to introduce noise into the more sensitive front end circuits.

Fender sure don't have problem when you put high gain into it. I did on two SF. It is sure a lot safer than a lot of the star grounding if people don't know what they are doing. I even gutted the Bassman and redo the circuit with tubes in new position by redrilling all the holes, I use the chassis as ground plane like Fender and it works perfect with cascade gain. So don't say it is not good for high gain. Yes, it is not the only way, but it is a good way AND a safe way.

Grounding IS A WHOLE LOT MORE THAN FACILITATE CONSTRUCTION. Do a research, as I said and I repeat, this is well written in EMC books and explain in detail. It has everything to do with hum and ground noise. Yes these amps are a lot simpler as it is not high gain, not low level signal. BUT people still runs into problems.

[jpfamps]

You need to force the high current path away from everything else.

If you re-read the grounding scheme I describe above you will notice that I do this, and it avoids any of the power supply ripple currents getting into the audio.

This scheme essentially is the same as that described in Douglas Self's Power Amplifier Design Handbook.

[Alan0354]

If you re-read the grounding scheme I describe above you will notice that I do this, and it avoids any of the power supply ripple currents getting into the audio.

This scheme essentially is the same as that described in Douglas Self's Power Amplifier Design Handbook.

Not if you are trying to tie the cathode of the power tubes to the grounding point of the screen filter cap after the choke!!! That's is just not right. You avoid this at all cost. You let the high noise power side isolated on an island. After the choke, then you use the chassis as ground plane for all the low current circuit. Also, have the two leads of the secondary of the OT connect to the speaker jack directly before to the chassis on the connector. This is to avoid the high current even touch the chassis.

[jpfamps]

Not if you are trying to tie the cathode of the power tubes to the grounding point of the screen filter cap after the choke!!! That's is just not right.

I will have disagree with you here.

The power valve grid reference (the ground end of the bias supply filter caps), the power valve cathodes and the filter cap for the screen supply should be tied to one node.

Any ground loop noise here induced onto the bottom of this node will be seem as a common mode signal and NOT be amplified.

Using the scheme above I've never had any problems with ground loops in amps (and I've constructed several from scratch, ie not kits, including high gain channel switching amps).

Now, it could be that I have been exceedingly lucky........

Merlin Blencowe also recomends this point of grounding of the screen supply filter cap and power valve cathode grounds.

See figure 13.13 in this below:

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

Merlin contributes to this forum (and has written two excellent books on valve amplifiers), so may like to comment on this.

[Alan0354]

I read page 8 of the article. I strongly disagree with the author. I scan through the page, I don't see any explanation on page 8. I see the rectifier large current on Page 4. But this is irrelevant in my design. This is my drawing:



I intentionally put the PA in an island by itself. You can see how the current complete the loop ( in Red)without even touching the ground. The resistor next to C is just the resistance of the lug of the star ground. But the current never even go into the chassis. In amps I've seen, the PI and screen voltage are the same stage which I draw as PI stage. The large current the article talked about is around the smaller loop from the rectifier to the line filtering cap. None of the heavy current gets dumped into the junction C to the chassis in the drawing. If there is no current from point C flowing into the chassis, Point C is not going to have any ripple in my drawing referencing to the chassis ground. That's the reason I specified a single point star at C.

If you look at the PI driver stage, as I explained that the current choose the path of least resistance or impedance depend on the frequency, as long as the loop in green color of the PI stage and it's filter cap is in close proximity, the loop will be very small even though it connect directly to the chassis. There should be no noise at point C, there should be no noise between point C and the ground connection of the PI stage.

BUT if you take the cathode from point A and connect to point B instead, then you dump the high current and noise from the power tubes into the quiet ground at point B of the PI section. Any noise at B will be amplified by the PI stage as common grid configuration through the cathode resistor through the cathode to the plate. Each half of the PI will have the gain of plate resistance divided by cathode resistor. For the PI in Fender, cathode resistor is about 22 to 33 K and some plate resistor is 100K, that is a noise gain of 4 or so. Yes, it is not that bad as it is also common mode noise. BUT the open loop gain of the PI PA section is not very high, common mode rejection is not as good as traditional op-amp. You definitely have more noise if you ground the cathode to the cap of the screen/PI stage.

As you can see, the way I make the preamp tube having a very small current loop between the tube and the cap and isolate by proximity effect on the chassis. If you force a one point star ground and you put the wrong stages together, the noise from one stage will be inject into the other stages on the same star. Now you are talking about big trouble. There will be no common mode noise rejection. Each stage of the preamp act as common grid stage that amplifier the injected noise through the cathode. Since you have bypass cap from the cathode to the ground, you inject to that point ground, the gain will be Rp/(1/gm) of the stage and it will be huge!!! Something like 60!!! In the drawing, I have resistance drawn in dashed lines and label "chassis", that is to show I have the filter cap of each stage close to the tube and the ground current goes in a loop. The two stages are farther apart and the chassis resistance is higher than the ground resistance within the loop of each stage, so not a lot of current cross over between stages.

You can really get into trouble if you are not careful putting the star ground. That's why it's so brilliant of Fender using a LOW INDUCTANCE ground strip along the length of the chassis and ground onto the metal chassis at regular intervals and make the chassis as a full ground plane. Since each tube is grounded along the length of the chassis, proximity effect isolate the stages.

As I have to stress one more time, ground current travel at the path of least resistance at very low frequency, travel at the path of least impedance at moderate and high frequency. Ground current in a ground plane do not travel in all direction as conventional believe. That is one of the major basis of EMC design. You can take advantage of the proximity effect and let the current choose it's path. Don't force the current into a star unless you know what you are doing. Making one critical star and make it simple like what I draw for the power amp. Notice I specified the low side of the speaker connect to ground by a small wire? That is to make sure no current from the secondary even touch the chassis.

[Steve Conner]

The currents and frequencies involved in tube amps are so low that these points are moot. JPF's, Merlin's and Alan's ground schemes will all work fine.

Just don't stuff up the wiring between rectifier and first filter cap, and ideally have a direct connection between the input jack sleeve and the chassis, to sink common-mode RF on the cable screen. I like to ground my preamp bus at the input jack, and do a "change of reference" at the balanced signal from the PI.

In RF and power electronic work, tiny changes in grounding can make huge differences. I've seen tens of volts developed across the inductance of a few inches of PCB trace. Using the wrong end of that as a reference can actually destroy gate driver ICs and the like.

In audio, you'll run up against compromises between ground loops, safety and EMC.

[jpfamps]

I read page 8 of the article. I strongly disagree with the author. I scan through the page, I don't see any explanation on page 8. I see the rectifier large current on Page 4. But this is irrelevant in my design. This is my drawing:

[ATTACH=CONFIG]20304[/ATTACH]

I intentionally put the PA in an island by itself. You can see how the current complete the loop ( in Red)without even touching the ground. The resistor next to C is just the resistance of the lug of the star ground. But the current never even go into the chassis. In amps I've seen, the PI and screen voltage are the same stage which I draw as PI stage. The large current the article talked about is around the smaller loop from the rectifier to the line filtering cap. None of the heavy current gets dumped into the junction C to the chassis in the drawing. If there is no current from point C flowing into the chassis, Point C is not going to have any ripple in my drawing referencing to the chassis ground. That's the reason I specified a single point star at C.

If you look at the PI driver stage, as I explained that the current choose the path of least resistance or impedance depend on the frequency, as long as the loop in green color of the PI stage and it's filter cap is in close proximity, the loop will be very small even though it connect directly to the chassis. There should be no noise at point C, there should be no noise between point C and the ground connection of the PI stage.

BUT if you take the cathode from point A and connect to point B instead, then you dump the high current and noise from the power tubes into the quiet ground at point B of the PI section. Any noise at B will be amplified by the PI stage as common grid configuration through the cathode resistor through the cathode to the plate. Each half of the PI will have the gain of plate resistance divided by cathode resistor. For the PI in Fender, cathode resistor is about 22 to 33 K and some plate resistor is 100K, that is a noise gain of 4 or so. Yes, it is not that bad as it is also common mode noise. BUT the open loop gain of the PI PA section is not very high, common mode rejection is not as good as traditional op-amp. You definitely have more noise if you ground the cathode to the cap of the screen/PI stage.

As you can see, the way I make the preamp tube having a very small current loop between the tube and the cap and isolate by proximity effect on the chassis. If you force a one point star ground and you put the wrong stages together, the noise from one stage will be inject into the other stages on the same star. Now you are talking about big trouble. There will be no common mode noise rejection. Each stage of the preamp act as common grid stage that amplifier the injected noise through the cathode. Since you have bypass cap from the cathode to the ground, you inject to that point ground, the gain will be Rp/(1/gm) of the stage and it will be huge!!! Something like 60!!! In the drawing, I have resistance drawn in dashed lines and label "chassis", that is to show I have the filter cap of each stage close to the tube and the ground current goes in a loop. The two stages are farther apart and the chassis resistance is higher than the ground resistance within the loop of each stage, so not a lot of current cross over between stages.

You can really get into trouble if you are not careful putting the star ground. That's why it's so brilliant of Fender using a LOW INDUCTANCE ground strip along the length of the chassis and ground onto the metal chassis at regular intervals and make the chassis as a full ground plane. Since each tube is grounded along the length of the chassis, proximity effect isolate the stages.

As I have to stress one more time, ground current travel at the path of least resistance at very low frequency, travel at the path of least impedance at moderate and high frequency. Ground current in a ground plane do not travel in all direction as conventional believe. That is one of the major basis of EMC design. You can take advantage of the proximity effect and let the current choose it's path. Don't force the current into a star unless you know what you are doing. Making one critical star and make it simple like what I draw for the power amp. Notice I specified the low side of the speaker connect to ground by a small wire? That is to make sure no current from the secondary even touch the chassis.

In your diagram B & C are connected together, via the chassis. In Merlin's diagram the equivalents of B & C are also connected together, but via a ground buss.

In your diagram there is no filter cap dedicated to the screen supply. It is customary to have a separate filter cap for the screen supply, and a dropping resistor before the filter cap to the PI.

You don't need to ground C and B separately, as the charging currents don't flow between B & C. The PA is still in an "island" in Merlin's scheme.

Best practice in audio these days is that the signal ground is connected to the chassis at one point near the input, and not use the chassis as a ground plane. This helps to avoid problems with ground loops when connecting to other equipment.

Your grounding of the transformer secondary also doesn't take into account application of negative feedback.

[R.G.]

It all reduces to this for grounding within one chassis:
1. all conductors are resistors, and generate a voltage when current flows through them
2. you need to know what currents flow through what paths
3. you need to select and arrange conductors to keep the voltages generated from the current flows in one path from being introduced into the signal chain in another path
4. you need to shield the signal path from both capacitive (electrical field) and inductive (magnetic field) radiated interference and feedback

Star grounding is not the only way to do it, but it is one way which is guaranteed to follow items 2 and 3. There are many other ways which can work, but which cannot necessarily be determined to do so ahead of time without modelling amounting to build-and-try.

Star grounding is so laborious that some modified version is usually used. Some rules of thumb for which places can be grouped into local stars have been developed.

Star grounding does not help you with item 4. You can still mess yourself up that way.

There are three common cases which are related to this which deserve mention.
1. The rectifier currents into and out of the first filter cap must be wired so that no part of the ground system shares any wires with the rectifier currents. Doing otherwise ignores item 3, and it will pass noise into your grounding system.
2. The rectifiers can generate RF pulses at the rectification frequency. These pulses are either conducted or radiated, and can be picked up and present themselves as "hum" in the circuit. This is really item 4, but the simplest cure is to snub the rectifiers to damp the RF or to change the rectifiers to something "self-snubbing".
3. If your power output (speakers, for audio amps) is referenced to signal ground, it needs to NOT use the chassis as a return path, nor share a conductor with a signal ground reference wires. The high currents in this wire cause voltages which can be reintroduced into the signal path. This would not be an issue for transformer isolated outputs like tube amps, except where the amplifier uses feedback from the transformer secondary, like many guitar amps do.

This is mostly a low frequency issue. Separate current paths are generally the best scheme at low frequencies. As frequencies get higher and the self-inductance of the wiring gets to be an issue, ground conductors cease being conductors and become transmission lines, where you need to go to what amounts to ground planes. Ground planes are not a panacea at low frequencies and may cause problems. They are mandatory for most RF work.

[Alan0354]

Here is a more detail comparison between mine and Merlin as I drew mine before I saw the Merlin article. The left side is mine, the right side is Merlin.



On mine, I added the heavy current loop from rectifier through the first cap as pointed out by Merlin. Then I draw the current path of the power tubes. As you can see, all current are contain inside the circuit, there is no current going out to the chassis ground. Thereby the voltage across the resistor from C to ground is zero due to the power amp stage. Only current through this lug resistance is the current from the preamp stages and PI which is low current. There is no way out of this one. There is no noise induced from the power section to the more sensitive front end. The PI being chassis ground reference will not amplify any noise from the rectifier and PA stage. In a lot of amps, the screen voltage pick off right after the choke. This is usually the same point that power the PI stage.

Now you look at Merlin assertion of tying point A to point B. Refer to the drawing on the right side, Now you look at the current path from the rectifier through the OT, through the power tube to point A. Then from point A to point B. Then the current has to travel from B back to the rectifier to complete the loop. See the problem? This current will develope noise voltage on the two chassis and stud resistance and will be amplified by the PI stage with gain of Rp1/Rc as shown. This is still ok if it is push pull PA where the same noise voltage is developed across both Rp1 and Rp2. This become common mode voltage to the PA and will be mostly cancel out.

BUT the Merlin scheme will fail if it is a SE power amp as there will be no common mode rejection to safe his day. He is wrong.

[Alan0354]

In your diagram B & C are connected together, via the chassis. In Merlin's diagram the equivalents of B & C are also connected together, but via a ground buss.
NO, point A connect to point C which is the lug of the star. Read my post about the path of least resistance and impedance. That is the whole point about signal traveling.
In your diagram there is no filter cap dedicated to the screen supply. It is customary to have a separate filter cap for the screen supply, and a dropping resistor before the filter cap to the PI.
This is irrelevant. This is a simple drawing, it will be exactly the same with a screen cap. The point is you don't want to introduce noise from PA to the rest of the circuit. In a lot of amps, the screen voltage is picked off after the choke. It is the same point where it power the PI. That's the reason I have the screen on the PI side.
You don't need to ground C and B separately, as the charging currents don't flow between B & C. The PA is still in an "island" in Merlin's scheme.
Not if he connect the cathode of the power tube to the screen and PI stage.
Best practice in audio these days is that the signal ground is connected to the chassis at one point near the input, and not use the chassis as a ground plane. This helps to avoid problems with ground loops when connecting to other equipment.
If you think of the ground current, how the current move, then the chassis ground plane is the best. The whole industry already move to full ground plane already for very good reason.
Your grounding of the transformer secondary also doesn't take into account application of negative feedback.
The secondary is floating by nature, the wire that tie one side of the secondary ONLY to establish a ground reference. That's the only way to make negative feedback possible. Without the ground reference, you can't even have a negative feedback.

I cannot write formulas here as it is not set up for that. If you want to get into more detail about ground current, these are the two posts that I answer and debate in Physics Forum that relate to ground current, signal flow and how to predict the current flow.

How does electricity really flow?
Any time varying signal travels as EM wave, not as charge or "current" as people think. Signal cannot depend on electrons moving as it is way too slow.

PCB layout question
I explained in post #14 and provide two link on how to predict current path. This is what I called proximity effect here.

This is the main part about EMC, that you can predict the current even in Audio. At this low frequency, you don't expect the image current on the ground plane follow right under the trace, but it will be something similar. The low frequency current still follow the path of least resistance.

[jpfamps]

Here is a more detail comparison between mine and Merlin as I drew mine before I saw the Merlin article. The left side is mine, the right side is Merlin.

[ATTACH=CONFIG]20311[/ATTACH]

On mine, I added the heavy current loop from rectifier through the first cap as pointed out by Merlin. Then I draw the current path of the power tubes. As you can see, all current are contain inside the circuit, there is no current going out to the chassis ground. Thereby the voltage across the resistor from C to ground is zero. There is no noise induced. The PI being chassis ground reference will not amplify any noise from the rectifier and PA stage. In a lot of amps, the screen voltage pick off right after the choke. This is usually the same point that power the PI stage.

Now you look at Merlin assertion of tying point A to point B. Refer to the drawing on the right side, Now you look at the current path from the rectifier through the OT, through the power tube to point A. Then from point A to point B. Then the current has to travel from B back to the rectifier to complete the loop. See the problem? This current will develope noise voltage on the two chassis and stud resistance and will be amplified by the PI stage with gain of Rp1/Rc as shown. This is still ok if it is push pull PA where the same noise voltage is developed across both Rp1 and Rp2. This become common mode voltage to the PA and will be mostly cancel out.

BUT the Merlin scheme will fail if it is a SE power amp as there will be no common mode rejection to safe his day. He is wrong.

I think you mis-represented Merlin's schematic.

A & B aren't connected as in the right-hand schematic you have drawn in Merlin's paper. The bottom of the bridge is connected directly to A and then A & B are connected together, thus the rectifier currents are contained and don't pass through B (as in your right hand schematic).

Regardless, Merlin contributes to this forum, so hopefully he will drop by and comment on this.

[Alan0354]

Fig 3.11 in page 8. He did not draw the PI tube which mostly are in the same circuit as the screen grid voltage right after the choke. Besides, why do you want to bring the noise out? You keep the highest current confine in an isolated island, not bring it out anywhere.

Let me emphasize again, you'll be in big trouble if this is a single end amp that do not provide any common mode rejection. You don't see this as a problem in push pull PA ONLY because the noise is mostly common mode appear in phase at the plates of the PI. Just like the ripple at the center tap of the OT do not couple to the speaker. With SE, there will be no common mode rejection and you'll run into problem.

I look forward to debate him if he is in this forum. I might not have years of experience with tube amps, but I am specialize in RF EM and EMC design........circuit with input in nV range and with gain of 80db+ and way into tens of MHz.

I can assure you, if you play with star grounding, you are playing with fire. This is not based on believe or mojo or black magic, this is pure theory.

[jpfamps]

Fig 3.11 in page 8. He did not draw the PI tube which mostly are in the same circuit as the screen grid voltage right after the choke. Besides, why do you want to bring the noise out? You keep the highest current confine in an isolated island, not bring it out anywhere.

Let me emphasize again, you'll be in big trouble if this is a single end amp that do not provide any common mode rejection. You don't see this as a problem in push pull PA ONLY because the noise is mostly common mode appear in phase at the plates of the PI. Just like the ripple at the center tap of the OT do not couple to the speaker. With SE, there will be no common mode rejection and you'll run into problem.

I look forward to debate him if he is in this forum. I might not have years of experience with tube amps, but I am specialize in RF EM and EMC design........circuit with input in nV range and with gain of 80db+ and way into tens of MHz.

I can assure you, if you play with star grounding, you are playing with fire. This is not based on believe or mojo or black magic, this is pure theory.....EM theory.

Merlin's schematic does confine the heavy power supply ripple currents, and they don't get into the audio (if you draw it correctly........).

I've used similar grounding schemes with single ended and push pull amps and reproducibly experience no significant hum.

Douglas Self (who know a bit about analogue audio........), also advocates a similar ground scheme.

Hopefully, Merlin will provide a more authoritative answer than I seem capable of supplying.

[Wombaticus]

Fig 3.11 in page 8. He did not draw the PI tube which mostly are in the same circuit as the screen grid voltage right after the choke. Besides, why do you want to bring the noise out? You keep the highest current confine in an isolated island, not bring it out anywhere.

Let me emphasize again, you'll be in big trouble if this is a single end amp that do not provide any common mode rejection. You don't see this as a problem in push pull PA ONLY because the noise is mostly common mode appear in phase at the plates of the PI. Just like the ripple at the center tap of the OT do not couple to the speaker. With SE, there will be no common mode rejection and you'll run into problem.

I look forward to debate him if he is in this forum. I might not have years of experience with tube amps, but I am specialize in RF EM and EMC design........circuit with input in nV range and with gain of 80db+ and way into tens of MHz.

I can assure you, if you play with star grounding, you are playing with fire. This is not based on believe or mojo or black magic, this is pure theory.

SE naturally don't have phase inverters, and in any case the conventional practice is to have a separate filter capacitor for the screen and for the PI. This is certainly the case with the Fender amps you refer to... to my knowledge, the larger Blackface and Silverface amps all had separate screen supply nodes. Even the tweed era 5E3 and 5F6-A Bassman had a dedicated filter cap for the screen. In fact, offhand, the only Fender amps that come to mind that DIDN'T have a dedicated filter cap for the screens was the very early 5E1 Champ and a few early Princetons. There may be others, sure, but it certainly was not the usual practice even in the 60s to have a shared PI/screen cap, and it is most definitely not the modern practice.

[Alan0354]

SE naturally don't have phase inverters, and in any case the conventional practice is to have a separate filter capacitor for the screen and for the PI. This is certainly the case with the Fender amps you refer to... to my knowledge, the larger Blackface and Silverface amps all had separate screen supply nodes. Even the tweed era 5E3 and 5F6-A Bassman had a dedicated filter cap for the screen. In fact, offhand, the only Fender amps that come to mind that DIDN'T have a dedicated filter cap for the screens was the very early 5E1 Champ and a few early Princetons. There may be others, sure, but it certainly was not the usual practice even in the 60s to have a shared PI/screen cap, and it is most definitely not the modern practice.

The point is you don't pull a high current path out and inject into the low noise high gain circuits. For SE, you don't have a PI, but if you connect the cathode from the PA to the screen drive that is on the other side of the choke, you inject noise into the low noise part of the circuit. There is no other way to put it.

If you don't pull the cathode of the power tube out, you don't need to have a filter just for the screen, you can use one cap for the screen and the PI. So why do you want to have a dedicate cap just for the screen, then go through a resistor before provide the HV for the PI? Just to put the cathode current into the screen circuit? for what?

Ultimately, why do you want to inject noise into the other part of the circuit if you don't have to?