This will also give you a safety redundancy for the HT ground. An open wire failure of the HT ground umbilical could give full HT between power supply and amplifier chassis' - very dangerous.
All good fortune,
Chris
Not sure if that is backwards or not, since I might be interpreting it incorrectly. But assuming that somewhere on the buss is the "zero volts" point, you don't want any currents flowing to the zero volts point that SHARE any of the buss with the signal ground reference's connection to the zero volts point.
For example, if you have points A, B, C, and D sequentially on a buss conductor, and B is the zero volts point, and D is the input signal reference ground conductor's connection to the buss, then nothing with any changing (OR large) current should be returning through C, since that current would then induce voltage between C and B that would then also be seen between D and B.
So it seems like the signal reference ground should be farthest "downstream", on the buss, assuming that "downstream" means closer to the zero volts point.
Not sure. (Sorry, I just got back from a funeral that was spread over three days and two cities that were very distant from here and from each other.).
First off, Gootee, thanks for the reply and sorry for your loss.
OK, so let's go with this example. If D is the input signal reference connection, and B is the zero volts point, and I have a wire with a large changing current to connect to the bus, where should I connect it for lowest noise? Point A would seem to be a good candidate as it does not share any of the bus with the signal reference ground's connection to B. Am I on the right track?
I didn't say charging current, I said changing current. I get the point about the charging current loops in the PSU, and have seen for my own ears that you were right about that.
To follow up on Gootee's example, if there is a wire with a changing current to connect to the bus (in other words a "noisy" wire), would the best place to connect that be to point A?
To follow up on Gootee's example, if there is a wire with a changing current to connect to the bus (in other words a "noisy" wire), would the best place to connect that be to point A?
Sorry, misread your post.
If you have a noisy wire to connect you need to think about where is the balancing current, because all currents must flow in loops. Connect the noisy wire as close as possible to where the opposite current comes in e.g. a valve cathode should connect near the anode decoupler.
If you have a noisy wire to connect you need to think about where is the balancing current, because all currents must flow in loops. Connect the noisy wire as close as possible to where the opposite current comes in e.g. a valve cathode should connect near the anode decoupler.
First, I have to say that I actually dislike the idea of buss grounding. It is not optimal, for avoiding external nor self-inflicted noise and interference. But yes, if you couldn't lay things out the best way, then point A would be better than point C, in the example.
Star grounding is the best "common" way, because it systematically minimizes the sharing of conductors by different and unrelated types of ground-return currents, minimizing the production of unwanted "backfed" voltages by currents from different and unrelated parts of the circuit. (But often, or usually, or maybe always, the best way would be to use a multilayer printed circuit board, with ground planes and power planes and signal planes, et al.)
Say your decoupling capacitors share a ground buss with your grid/base/gate input resistor's grounded end. Even with no signal, any AC ripple on the power rails that goes through the capacitors and the ground buss would induce a voltage across the inductance and resistance of the buss conductor. That voltage (or a voltage caused by it) would appear back at the upstream end of every conductor connected on that part of the buss. So the "ground" end of the input resistor would then have a time-varying AC voltage, instead of a steady (near-)zero volts. That "bouncing ground" voltage would in most cases simply sum itself, arithmeticaly, with the voltage seen by the signal input pins of the amplification device. That would be "a BAD thing". (Add a signal or something else that that disturbs the rails and you have an even worse problem. And that's from only one of the many possible types of sources of shared ground-return currents.)
As long as I'm already standing on this soapbox, switching gears somewhat, all "enclosed loop areas" should be minimized, to the extreme if possible. Just like with input signal and ground, all other "related pairs" (which form closed loops, topologically, actually) should always run as close together as possible. Any loop that encloses geometric area will have time-varying currents induced in it, by any time-varying electromagnetic fields in the air. And any time-varying currents in a loop will produce time-varying electromagnetic fields in the air. So you can unwittingly have both receivers and/or transmitters. The relative amplitudes of the induced currents and EM fields will be proportional to the total amount of geometric area enclosed by a loop, all else being equal.
Both cases (receiving and transmitting loops) can be very bad. Imagine separating the input signal and ground conductors, so there is space between them (i.e. creating "enclosed loop area"). The induced loop currents, from, say, the AC mains fields in the air, would induce voltages across every impedance in the loop, including the resistor across the amplification device's input pins: Hum! RF would get in, that way, too. So not only should the input pairs be shielded twisted pair, or twisted pair, or stacked PCB traces, so should the AC Mains pair, and the rectified AC and ground pair, and each DC power and ground pair, and the output pairs. If you're not going to shield a pair of wires, then at least twist them tightly together, with five or six turns per inch, ALL the way to each end.
If you see any conductor routed alone, you should wonder why it's not mated as part of a pair and think about how to best do that. Sometimes, unfortunately, power and ground planes are almost a necessity, for making it practical to eliminate most enclosed loop area.
In the case of an umbilical, I think that probably everything should be paired.
Star grounding is the best "common" way, because it systematically minimizes the sharing of conductors by different and unrelated types of ground-return currents, minimizing the production of unwanted "backfed" voltages by currents from different and unrelated parts of the circuit. (But often, or usually, or maybe always, the best way would be to use a multilayer printed circuit board, with ground planes and power planes and signal planes, et al.)
Say your decoupling capacitors share a ground buss with your grid/base/gate input resistor's grounded end. Even with no signal, any AC ripple on the power rails that goes through the capacitors and the ground buss would induce a voltage across the inductance and resistance of the buss conductor. That voltage (or a voltage caused by it) would appear back at the upstream end of every conductor connected on that part of the buss. So the "ground" end of the input resistor would then have a time-varying AC voltage, instead of a steady (near-)zero volts. That "bouncing ground" voltage would in most cases simply sum itself, arithmeticaly, with the voltage seen by the signal input pins of the amplification device. That would be "a BAD thing". (Add a signal or something else that that disturbs the rails and you have an even worse problem. And that's from only one of the many possible types of sources of shared ground-return currents.)
As long as I'm already standing on this soapbox, switching gears somewhat, all "enclosed loop areas" should be minimized, to the extreme if possible. Just like with input signal and ground, all other "related pairs" (which form closed loops, topologically, actually) should always run as close together as possible. Any loop that encloses geometric area will have time-varying currents induced in it, by any time-varying electromagnetic fields in the air. And any time-varying currents in a loop will produce time-varying electromagnetic fields in the air. So you can unwittingly have both receivers and/or transmitters. The relative amplitudes of the induced currents and EM fields will be proportional to the total amount of geometric area enclosed by a loop, all else being equal.
Both cases (receiving and transmitting loops) can be very bad. Imagine separating the input signal and ground conductors, so there is space between them (i.e. creating "enclosed loop area"). The induced loop currents, from, say, the AC mains fields in the air, would induce voltages across every impedance in the loop, including the resistor across the amplification device's input pins: Hum! RF would get in, that way, too. So not only should the input pairs be shielded twisted pair, or twisted pair, or stacked PCB traces, so should the AC Mains pair, and the rectified AC and ground pair, and each DC power and ground pair, and the output pairs. If you're not going to shield a pair of wires, then at least twist them tightly together, with five or six turns per inch, ALL the way to each end.
If you see any conductor routed alone, you should wonder why it's not mated as part of a pair and think about how to best do that. Sometimes, unfortunately, power and ground planes are almost a necessity, for making it practical to eliminate most enclosed loop area.
In the case of an umbilical, I think that probably everything should be paired.
Last edited:
I built a spud 5842 SET headamp with a separate power supply for each channel and a separate chassis for the power supply. This is for a headphone with 97db per milli watt. This requires a tremendously quiet amp.
The most important advice is the power supply filter ground return going to the CT, this can be the final star ground. But you want to get that return loop close and tight. In the amp section you can have two star ground "nodes" that lead back to the CT in the PS section. I don't agree with leaving the signal input ground "floating" it should be tied to the star ground in the amp chassis. The filaments are going to be your biggest problem, I don't think many have experience with a two chassis DC filament DHT preamp design, its in developmental stage IMHO.
But when you get it right, the biggest contributor to 60hz hum with be radiation from the B+ transformers to the OPT's, you will watch the 60hz peak drop as you separate the two chassis.
The most important advice is the power supply filter ground return going to the CT, this can be the final star ground. But you want to get that return loop close and tight. In the amp section you can have two star ground "nodes" that lead back to the CT in the PS section. I don't agree with leaving the signal input ground "floating" it should be tied to the star ground in the amp chassis. The filaments are going to be your biggest problem, I don't think many have experience with a two chassis DC filament DHT preamp design, its in developmental stage IMHO.
But when you get it right, the biggest contributor to 60hz hum with be radiation from the B+ transformers to the OPT's, you will watch the 60hz peak drop as you separate the two chassis.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.