My two penny's worth:-
What he is saying is that if there are more than two connections to earth(ground) in a system of components (e.g. source, preamp and amp) you will in all likely hood create an earth/ground loop which is quite right.
The issue that will be raised (legitimately) with only having 1 earth point in a complete system like this is safety.
In commercial gear, you get around it by producing equipment that is double insulated (DI) and there are no earth connections e.g. most CD players have no earth, and many integrated amps from Japan etc are also double insulated.
But, if you cannot do this, and I don't think DIYers should be trying to create DI stuff because one mistake and you are in trouble, you are left with Ground Lifters in each piece of gear. Do it on one piece of gear, but to do it in a pre and then a power amp piles the risks up IMV. I talk about ground lifters in my presentation but recommend you use a 35A/400V bridge.
In my commercial gear, everything is earthed - there are no ground lifters, but there is the option to use balanced interconnects from pre to power and that would address the noise issue - but the stuff is still quiet on unbalanced.
Re the earth connection sequence you quoted, again on that point you have to be careful. If I am using a ground lifter (as in the sx and nx amplifiers) I still connect the incoming mains earth(ground) directly to the chassis for safety purposes and then from the grounded chassis to the PSU 0V, I connect the ground lifter.
The problem I have with taking the safety earth(ground) to the component 0V and then from there to the chassis via a ground lifter is safety. Lose the GL and have a live onto the electronics and you will have a safety problem. Would this get through UL certification?The safety earth(ground) is the most important connection in the amp.
I put a small (5-10nF) ceramic across the ground lifter bridge so that at RF it is shorted out. This won't get around the low frequency earth(ground) voltage differential coupling capacitively to everything in the box that ilimzn mentions, but I have not found a problem in that regard in any of my tests, but then again the wiring practices and regulations in the UK are pretty strict - ditto rest of the EU and the USA so these kinds of problems are pretty rare. Also note that if all the electronics in the system are connected to earth(ground) via their 0V, the chassis and the electronics wont be experiencing the issue anyway since everything is at the same potential.
What he is saying is that if there are more than two connections to earth(ground) in a system of components (e.g. source, preamp and amp) you will in all likely hood create an earth/ground loop which is quite right.
The issue that will be raised (legitimately) with only having 1 earth point in a complete system like this is safety.
In commercial gear, you get around it by producing equipment that is double insulated (DI) and there are no earth connections e.g. most CD players have no earth, and many integrated amps from Japan etc are also double insulated.
But, if you cannot do this, and I don't think DIYers should be trying to create DI stuff because one mistake and you are in trouble, you are left with Ground Lifters in each piece of gear. Do it on one piece of gear, but to do it in a pre and then a power amp piles the risks up IMV. I talk about ground lifters in my presentation but recommend you use a 35A/400V bridge.
In my commercial gear, everything is earthed - there are no ground lifters, but there is the option to use balanced interconnects from pre to power and that would address the noise issue - but the stuff is still quiet on unbalanced.
Re the earth connection sequence you quoted, again on that point you have to be careful. If I am using a ground lifter (as in the sx and nx amplifiers) I still connect the incoming mains earth(ground) directly to the chassis for safety purposes and then from the grounded chassis to the PSU 0V, I connect the ground lifter.
The problem I have with taking the safety earth(ground) to the component 0V and then from there to the chassis via a ground lifter is safety. Lose the GL and have a live onto the electronics and you will have a safety problem. Would this get through UL certification?The safety earth(ground) is the most important connection in the amp.
I put a small (5-10nF) ceramic across the ground lifter bridge so that at RF it is shorted out. This won't get around the low frequency earth(ground) voltage differential coupling capacitively to everything in the box that ilimzn mentions, but I have not found a problem in that regard in any of my tests, but then again the wiring practices and regulations in the UK are pretty strict - ditto rest of the EU and the USA so these kinds of problems are pretty rare. Also note that if all the electronics in the system are connected to earth(ground) via their 0V, the chassis and the electronics wont be experiencing the issue anyway since everything is at the same potential.
Last edited:
Agreed Andrew. His posts were very interesting and thought provoking. Ideally the circuit 0V would be connected directly to the enclosure. I don't know enough about lifting that from safety earth but he seems pretty certain it's not an issue.
Of course it's only a problem if loops are created, so, since we are DIYing let's forget what we might have to do if we were making a commercial product and had no control over it's connection, the logical thing then would be to daisy chain the safety earth to our separate pieces of equipment.
Of course it's only a problem if loops are created, so, since we are DIYing let's forget what we might have to do if we were making a commercial product and had no control over it's connection, the logical thing then would be to daisy chain the safety earth to our separate pieces of equipment.
What he is saying is 0V to chassis then to PE via GL, not that it makes much difference to the safety issue that concerns you if the GL fails?
I am not happy with that. If the GL is faulty or fails? wire off? GL opens up? At least with a direct earth(ground) connection to the chassis you have a better chance.
IIRC the rules around how the PE connection is made are very strict - e.g. a closed circular lug bolted to the chassis using a serrated washed and a nut and then any other connections to earth on top of this using a separate second nut - i.e. you cannot pile the connections on top of one another and use a single nut.
IIRC the rules around how the PE connection is made are very strict - e.g. a closed circular lug bolted to the chassis using a serrated washed and a nut and then any other connections to earth on top of this using a separate second nut - i.e. you cannot pile the connections on top of one another and use a single nut.
"In my commercial gear, everything is earthed - there are no ground lifters, but there is the option to use balanced interconnects from pre to power and that would address the noise issue - but the stuff is still quiet on unbalanced."
Correction - the phono pre has a GL using 2 100 Amp diodes back to back MUR860
Correction - the phono pre has a GL using 2 100 Amp diodes back to back MUR860
Attached is the circuit analysis that demonstrates how R4 (ref figure of post #119) is outside the feedback loop. If you (or anyone else) believe any of this to be incorrect, please could you point out which of the equations 1 through 7 is in error?
Reposted from Bobs book thread by request:
The power amplifiers really need to be broken into small signal and output transistors.
Let's look at the current path for the speaker.
Current starts at the power supply, runs though a wire and maybe a rail fuse, then through a power transistor, emitter resistor, and another wire to the speaker terminal. After returning from the speaker, it needs to get back to the power supply ground to complete the circuit. This is a path with high intermittent peak currents and no part of this path should be shared with anything else.
The small signal portion of the power amplifier is relatively constant current and should be connected with separate wires back to the power supply to avoid contamination from the high current speaker loop. The input ground can then be connected either to the small signal portion of the amplifier or the star ground since the connecting ground wire will be relatively quiet.
The power amplifiers really need to be broken into small signal and output transistors.
Let's look at the current path for the speaker.
Current starts at the power supply, runs though a wire and maybe a rail fuse, then through a power transistor, emitter resistor, and another wire to the speaker terminal. After returning from the speaker, it needs to get back to the power supply ground to complete the circuit. This is a path with high intermittent peak currents and no part of this path should be shared with anything else.
The small signal portion of the power amplifier is relatively constant current and should be connected with separate wires back to the power supply to avoid contamination from the high current speaker loop. The input ground can then be connected either to the small signal portion of the amplifier or the star ground since the connecting ground wire will be relatively quiet.
New vs. Vintage, Power amp vs. integrated or receiver.
It appears that most of those on this thread are building new, stand alone, power amplifiers where one can start with an empty box, position the components for best response, and install the correct safety ground.
In my case, it is restoration and modernizing a vintage receiver that was designed 55 years ago. There is no possibility of relocating the transformer, output transistors, rail fuses, output terminals, input jacks, or speaker selector switch and associated wiring. The small signal section of the power amplifiers is limited as to possible locations. Cable routing has certain limitations also. There is also a preamp and tuner to deal with and some parts of those have fixed locations. So, one must simply do the best possible restoration and accept less than perfect results.
Additionally, the line cord is 2 wire, non polarized, and will not be replaced. The only chassis connection was the line capacitor and a resistor. The capacitor is now replaced with a proper XY rated capacitor and is connected line to line, rather than line to chassis. This leaves no connection to chassis other than capacitive coupling in the power transformer. The non polarized plug has been marked as to which side should go to neutral to have the lesser amount of leakage. However, it is well within limits either way. There should be an additional large value resistor (820K Ω is common) from chassis to one side of the line of anything having an external antenna connection to dissipate possible static charges.
A new design being sold must meet all of the present requirements but vintage only needs to meet what was in effect when it was originally sold. One can still buy an old AC/DC hot chassis radio with non polarized plug from 70 years ago and it is legal.
It appears that most of those on this thread are building new, stand alone, power amplifiers where one can start with an empty box, position the components for best response, and install the correct safety ground.
In my case, it is restoration and modernizing a vintage receiver that was designed 55 years ago. There is no possibility of relocating the transformer, output transistors, rail fuses, output terminals, input jacks, or speaker selector switch and associated wiring. The small signal section of the power amplifiers is limited as to possible locations. Cable routing has certain limitations also. There is also a preamp and tuner to deal with and some parts of those have fixed locations. So, one must simply do the best possible restoration and accept less than perfect results.
Additionally, the line cord is 2 wire, non polarized, and will not be replaced. The only chassis connection was the line capacitor and a resistor. The capacitor is now replaced with a proper XY rated capacitor and is connected line to line, rather than line to chassis. This leaves no connection to chassis other than capacitive coupling in the power transformer. The non polarized plug has been marked as to which side should go to neutral to have the lesser amount of leakage. However, it is well within limits either way. There should be an additional large value resistor (820K Ω is common) from chassis to one side of the line of anything having an external antenna connection to dissipate possible static charges.
A new design being sold must meet all of the present requirements but vintage only needs to meet what was in effect when it was originally sold. One can still buy an old AC/DC hot chassis radio with non polarized plug from 70 years ago and it is legal.
To simplify things, I have set the input voltage to 0V(and the output voltage is therefore also 0V).
The error voltage between power and amp grounds is 1mV, could be any amount, it is just to demonstrate what is taking place.
You will see a current flow through the load, through the amp, back to the PSU to force the output back to 0V. How else do you get 0V at the output?
Last edited:
1. What we want is for the voltage developed across the loudspeaker to be an exact non-distorted multiple of the input voltage.
2. So with an input of zero, the voltage developed across the loudspeaker should be zero.
3. In your example, we have an input of zero, but a voltage across the loudspeaker of -1 mV.
4. How has this happened? It's because current flowing in R4 has developed a 1 mV error voltage...
5. Which the feedback loop cannot correct, because R4 is outside the feedback loop.
Which of these statements do you disagree with, if any? And, which of the equations 1 through 7 in the circuit analysis I posted is not correct?
2. So with an input of zero, the voltage developed across the loudspeaker should be zero.
3. In your example, we have an input of zero, but a voltage across the loudspeaker of -1 mV.
4. How has this happened? It's because current flowing in R4 has developed a 1 mV error voltage...
5. Which the feedback loop cannot correct, because R4 is outside the feedback loop.
Which of these statements do you disagree with, if any? And, which of the equations 1 through 7 in the circuit analysis I posted is not correct?
There is another way to look at this.
The amp keeps the differential input voltage at zero, minus the non-zero error of the feedback loop.
If the input ground is at +1mV, then signal input is also at +1mV, and so the amp will change the output to keep the negative input at +1mV as well. Since input ground is at 1mV and the negative input must be at 1mV, there must be zero voltage across the lower feedback resistor. The only way to satisfy this is if there is also zero voltage across the upper feedback resistor, as any current flowing through this resistor will also flow through the lower feedback resistor and cause a voltage across it. So the output must also be at 1mV.
The amp keeps the differential input voltage at zero, minus the non-zero error of the feedback loop.
If the input ground is at +1mV, then signal input is also at +1mV, and so the amp will change the output to keep the negative input at +1mV as well. Since input ground is at 1mV and the negative input must be at 1mV, there must be zero voltage across the lower feedback resistor. The only way to satisfy this is if there is also zero voltage across the upper feedback resistor, as any current flowing through this resistor will also flow through the lower feedback resistor and cause a voltage across it. So the output must also be at 1mV.
All analysis and argument thus far has been assuming an ideal amplifier with infinite open-loop gain. Let's not complicate things further by trying to consider a finite-gain amplifier with non-zero feedback error.
Relative to which node?
Relative to which node?
It would really help if you could clarify whether or not you agree with me, and if not, in which equation of my analysis have I made a mistake?
The feedback loop is from the top of R2 to the bottom of R3. The feedback loop does not care what is going on to the the right of that ie the voltage drops etc across the 0V umbilical.
If you measured from the top of R2 to the bottom of R3 while injecting noise through the 0V return you would see no change in an ideal situation.
If you measured across the speaker, you would see the aforementioned voltage minus the error voltage across the 0 V umbilical.
If you wanted to include the 0V umbilical error, you would have to take the bottom of R3 to the right hand side of it.
Harry is correct on this!
If you measured from the top of R2 to the bottom of R3 while injecting noise through the 0V return you would see no change in an ideal situation.
If you measured across the speaker, you would see the aforementioned voltage minus the error voltage across the 0 V umbilical.
If you wanted to include the 0V umbilical error, you would have to take the bottom of R3 to the right hand side of it.
Harry is correct on this!
You will not see a voltage change over R2-R3. You will see a change in output current. Could someone please simulate this.
So you agree that the voltage over the speaker is not what it should be. It is output voltage minus the voltage between PSU and amp. And, that the current is not zero.
So you agree that the voltage over the speaker is not what it should be. It is output voltage minus the voltage between PSU and amp. And, that the current is not zero.
Last edited:
I think that the issue is directly relevant to how to wire an amplifier, no? i.e. it shows that it is vital that R4 is as small as possible, if you take the loudspeaker return to the PSU 0 V instead of AMP 0 V.
Having said that, I think that we probably don't need any more to argue the semantics of what is inside and outside of the feedback loop.
Not my point at all. Previously the we were only looking at the bad of one method, as if the other option doesn't have serious drawbacks. Clearly you would not want to connect the speaker return to PSU-0 if you're building a single channel amp. Clearly as can be with something as complex as grounding.
Can't wait untill we move on to the causes of hum.
Can't wait untill we move on to the causes of hum.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.