"Derived ground" for lack of a better term-
I first saw this trick in a Krohn-Hite tube function generator from the early '50s. They needed dc coupled outputs to generate low frequiencies and multiple regulated DC supplies in the tube days would be very expensive. So they made a single conventional regulated supply, floated both ends and put a "center tap regulator" on it to create a ground reference point for the supplies.
I first combined that concept with a current source supply and shunt regulators on the ENTEC Number Cruncher wat too many years ago. Bob Crump (of Blowtorch note) modified that by playing with the transformer polarities on the supplies (it used 3 transformers etc, for the various supplies). I didn't really buy into his transformer polarity, since there was a side effect of more radiated fields as the center transformer was flipped but its a very small point.
I'm using a similar supply architecture in the NuForce P9 (and the upcoming G9) so that there is no ground connection between chassis except the shields. Each amp floats except for the single ground connection that the input, output and center of the shunt regulators reference to.
The goal of this is to minimize ground currents. I try to balance all of the DC currents so there is no DC flowing back to ground. its never perfect but it seems to just help both in lower measured noise and better sound.
I first saw this trick in a Krohn-Hite tube function generator from the early '50s. They needed dc coupled outputs to generate low frequiencies and multiple regulated DC supplies in the tube days would be very expensive. So they made a single conventional regulated supply, floated both ends and put a "center tap regulator" on it to create a ground reference point for the supplies.
I first combined that concept with a current source supply and shunt regulators on the ENTEC Number Cruncher wat too many years ago. Bob Crump (of Blowtorch note) modified that by playing with the transformer polarities on the supplies (it used 3 transformers etc, for the various supplies). I didn't really buy into his transformer polarity, since there was a side effect of more radiated fields as the center transformer was flipped but its a very small point.
I'm using a similar supply architecture in the NuForce P9 (and the upcoming G9) so that there is no ground connection between chassis except the shields. Each amp floats except for the single ground connection that the input, output and center of the shunt regulators reference to.
The goal of this is to minimize ground currents. I try to balance all of the DC currents so there is no DC flowing back to ground. its never perfect but it seems to just help both in lower measured noise and better sound.
john curl said:
Not understandable to me. This is the task I would never delegate to anyone else.
JPV said:
Sliding in here, I thought this was touched on already a few hundred pages ago.
My simple reduction-
1) Reed relays have a magnetic element in series with the signal (the reed). Very possibly impacting the sound.
2) Contact materials
For low level signals gold is the best for two reasons, first it resists corrosion better than other materials, second, it microwelds with less energy, which is how the contact works.
Mercury would be better still, with lower contact resistance faster response etc. but its a hazardous material and only used in reed relays as far as I can tell (excluding contactors).
Silver is better for high currents with lower resistance and longer life, but it takes more energy to reliably conduct or a significant wiping action.
Tungsten is good for high energy applications with very long life, but has a higher contact resistance
3) Contact styles Look for crossbar contacts, they have a wiping action that cleans the surface as they connect. Used a lot in telecommunications systems for high reliability.
4) Smaller is better, fewer parasitics to deal with. The Blowtorch used the TO5 can relays I believe (Bob Crump told me). These are great, except very pricey - about $25 ea. last I looked. They are common in avionics and RF/Microwave stuff.
Others may have different experiences and opinions.
CTC is the combined initials of 3 separate designers. Bob Crump, Carl Thompson, and me.
Bob built virtually all the units, and he had more experience with grounding that I did. Carl did the board layout, and I designed and built the line amp, phono, and power supply boards. IF we had some real problems with grounding, then I would have been consulted, in order to help.
Bob built virtually all the units, and he had more experience with grounding that I did. Carl did the board layout, and I designed and built the line amp, phono, and power supply boards. IF we had some real problems with grounding, then I would have been consulted, in order to help.
John:
Sorry, I stand corrected. The big silver on silver switches are the best option where possible. It's been a long time and I don't remember the context of the discussion with Bob.
Sorry, I stand corrected. The big silver on silver switches are the best option where possible. It's been a long time and I don't remember the context of the discussion with Bob.
1audio said:
Interesting. I'm assuming that Krohn-Hite was shifting the power supply reference relative to the true circuit ground to obtain a DC coupled output.
From the description of your implementation it's not clear if you are you using an active derived ground to track and maintain a zero reference, or just controlling or minimizing the ability of the the circuit bias currents from contaminating the signal reference. But it does sounds like your aware of the same ills I'm concerned with.
My approach has been to identify each current loop, it's source and return, to order them by their magnitude, and make layout decisions based on this. At least that's part of it. Sound complicated but it's more of a frame of mind that I'm trying to describe with words.
You are right though, when the signal references are clean their is a noticable improvement in the sound. Although there are probably many ways to effect the same results.
Regards, Mike.
1audio said:
Thinking it through a little more deeply I think I see what your doing. The input, output and regulator grounds are all still referenced to the raw supply ground and the derived ground is used to supply the reference to the input stage and FB.
If not, oh well, just more food for thought.
Regards, Mike.
Hi,
Nakamichi used a novel approach (to me) for local regulation. They worked to keep any currents returned to ground as small as possible.
For example, the ground connection from the regulator circuit went to the emitter of a transistor. The collector was connected to the opposite "raw" regulated DC line. The base of this transistor was connected to ground. This reduced the reference currents to ground by the beta of the transistor and passed the bulk of the current to the opposite supply before the other regulator.
The end result was a very quiet ground and it made a difference that was audible. This was also a lesson well learned as soon as I grasped what they were doing.
Now a question on signal relays. Why not use a DC bias (from a very clean source) to establish the audio path? This would eliminate capacitors and allow a single supply to be used. A "ground reference" could be used for common connections. Either that or build the circuit as normal and pass current to ground or the negative supply. I favour sending current into a ground reference over involving a supply rail.
Just a thought, not necessarily useful or practical. This is why telecom circuitry does this though, to minimize the effect of less than perfect contacts.
-Chris
Nakamichi used a novel approach (to me) for local regulation. They worked to keep any currents returned to ground as small as possible.
For example, the ground connection from the regulator circuit went to the emitter of a transistor. The collector was connected to the opposite "raw" regulated DC line. The base of this transistor was connected to ground. This reduced the reference currents to ground by the beta of the transistor and passed the bulk of the current to the opposite supply before the other regulator.
The end result was a very quiet ground and it made a difference that was audible. This was also a lesson well learned as soon as I grasped what they were doing.
Now a question on signal relays. Why not use a DC bias (from a very clean source) to establish the audio path? This would eliminate capacitors and allow a single supply to be used. A "ground reference" could be used for common connections. Either that or build the circuit as normal and pass current to ground or the negative supply. I favour sending current into a ground reference over involving a supply rail.
Just a thought, not necessarily useful or practical. This is why telecom circuitry does this though, to minimize the effect of less than perfect contacts.
-Chris
anatech said:
Chris.
That is a trick I have often used in my (simulated) amplifier designs.
Sometimes to minimize current return to GND.
But may as well be to refer to some other voltage level point in amp.
Like output etc.
As a matter of fact, I used this approach in one 2SK389 input and Exicon N-Channel Devices output amp.
The simulated results were so good, that I do not want to publish this actual schematic of this idea. Not yet.
Let's see if in future 😉
Notice that even this very Original Idea by me
Opto Power Amplifier - 100% NPN
http://www.diyaudio.com/forums/showthread.php?threadid=131822 /// Opto Power NPN - schematic
.... deals with avoiding 'rails currents' effects and to isolate Current Loops from eachother.
Like what MikeBettinger talks about. I have no problem understand Mike, in this case.
Regards, Lineup
Hi lineup,
Interesting circuit. Optocouplers are very non-linear and many can be as low as 20% current transfer. You would have to restrict the current swing in these too, or your feedback might no be able to compensate or the characteristics of the optocouplers. The encased transistor will also suffer from the same Early Voltage effects and the Miller capacitance might be higher. Just guessing, but these features are of little concern to designers using these products.
Still, it's certainly thinking outside the box. Wishing you success developing your amp. It will be the "LCOA" design. Light Coupled Output Activator. Either that or "BB", Beam Blaster for the disco crowd.
-Chris
Interesting circuit. Optocouplers are very non-linear and many can be as low as 20% current transfer. You would have to restrict the current swing in these too, or your feedback might no be able to compensate or the characteristics of the optocouplers. The encased transistor will also suffer from the same Early Voltage effects and the Miller capacitance might be higher. Just guessing, but these features are of little concern to designers using these products.
Still, it's certainly thinking outside the box. Wishing you success developing your amp. It will be the "LCOA" design. Light Coupled Output Activator. Either that or "BB", Beam Blaster for the disco crowd.
-Chris
MikeBettinger said:
Not quite but close. The supply floats and is external to the box. It's a current source with as high an impedance as I can make it. Internal to the box I further isolate it with resistors. The stacked shunt regulators define the regulated voltages. With luck there is very little dc going to the center of the stacked shunt regulators. The local supply caps should be large enough to support the peak current requirements of the load.
lineup said:
The optically coupled amp is something I think I saw years ago. There may be some benefit but the delay through the couplers and the non-linearity that needs a resistive connection to fix may undo all of the good.
I built a power amp once that only connected the driver to the output through the gates of the output fets and the feedback resistors. Being balanced, with a bridge tied load there was no need for a ground connection between the input and the output. There were no connections between the supplies. It worked pretty well but the output supplies would not settle balanced. I finally used a pair of resistors tied back to the input ground to lock the output supply down.
anatech said:
Hi Chris,
Maybe I don't fully understand but I would think that referencing the regulated supply to the opposite polarity raw supply would have it's own issues. This would be interesting to try.
Is there a Nakamichi model that I could see an example of this in?
The AES abstract posted yesterday is also a very interesting concept.
Regards, Mike.
1audio said:
Demian,
Assuming that I use a reed that is hermetically sealed, does that take care of the corrosion? Or is there still some sort of deterioration possible in vacuum?
I also once used reeds with sputtered ruthenium contacts, supposedly very good for low-level signals. Do you have any experience with those?
Jan Didden
tvi said:
Dead link? Got pulled??
_-_-bear
PS. iirc there was some discussion somewhere that included a comparison of advanced regulators vs. batteries and it was determined that the batteries had more self noise than the best regulators...
PPS. FYI: http://www.ne.jp/asahi/sound.system/pract/eparts/34degawa-final.pdf
janneman said:
Two things- first there are a lot of noble metal options and they all work to some degree. I don't know who would be "best". The best stuff I have read was from Gordos, but they have been bought and destroyed. . . This is pretty good explanitory stuff: Tyco Relay
Inside the sealed tube external contaminants won't get in BUT, like a vacuum tube, the internals will outgas and the contact arcing will generate the pollutants. There is no free lunch.
1audio said:
As you can not provide even a conceptual diagram
I can not comment on if yuor idea is good or not.
My unique and original idea with OPTO Couplers is very good.
I am sure.
It is more a matter of time
before getting the correct tools to make a great application for my concept.
As you read in my first post,
I mention the shortage, right now in time,
of higher speed good linearity OPTOS.
But I am not convinced using the best analog OPTOS of today
would render one Amplifier that is not very good for the Audio Band.
Notice that even this very Original Idea by me
.... deals with avoiding 'rails currents' effects
and to isolate Current Loops from eachother.
Opto Power Amplifier - 100% NPN
http://www.diyaudio.com/forums/showthread.php?threadid=131822
Regards, Lineup
.... deals with avoiding 'rails currents' effects
and to isolate Current Loops from eachother.
Opto Power Amplifier - 100% NPN
http://www.diyaudio.com/forums/showthread.php?threadid=131822
Regards, Lineup
I thought you could get opto coupled transistors- essentially power mosFETs with photosensitive gates.
I'm not sure how your design is less sensitive to ground currents. And getting the isolation for the floating 10V supplies is a challenge unto itself. The DC-DC converters have some capacitance in to out and lots of noise.
I'm not sure its worth going into more details on the first Spectral amp (the CPU 1). Its ancient history.
I'm not sure how your design is less sensitive to ground currents. And getting the isolation for the floating 10V supplies is a challenge unto itself. The DC-DC converters have some capacitance in to out and lots of noise.
I'm not sure its worth going into more details on the first Spectral amp (the CPU 1). Its ancient history.
- Status
- Not open for further replies.