checked the Meta, still have some ?s-- good Power Supply layout practices

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

soapfoot

Well-known member
Joined
Dec 27, 2010
Messages
1,065
Location
LA + Brooklyn
I was reading around in the Meta trying to gain some insight into good layout practices for laying out quiet audio power supplies.  I'm particularly interested in tube circuits.  I ended up raising more questions than answers.

In particular, on this thread: http://www.groupdiy.com/index.php?topic=22780 there was a quote from John Roberts which made me very curious for specifics.  He says:

In unregulated circuits this ripple can get into the audio via PSR, so in principle reducing ripple voltage could reduce noise floor, but as often as not ground contamination from charging current and PS layout will dominate, so as always, life is a simultaneous equation with many parts. Optimize one at the expense of another.

I'd like to know more about this, if possible, as well as anything else that could help me squeeze the last few dB of quiet out of power supply designs.  

Again from reading the meta, I understand that the concept of "ground" can be misleading, because there can be many return paths that you might not necessarily want together.  I see a lot of people advocate star grounding, but I see others maintain that certain high-current grounds should be located away from low-level audio grounds.  I'd love to know more about what types of grounds should be grouped together, and what types should be kept separate-- and how they should be kept separate.

Another question would be-- what is the functional difference between two chassis grounds some distance apart on an aluminum chassis, and connecting those two points with wire or buss bar in a star-grounding scheme?

Also, the phrase "ground contamination from charging current" is interesting.  I take this to mean that large charging capacitors can cause contamination of ground?  I'd like to know more. 

Thanks for any insight at all.  There's a lot about this that's mysterious to me.

 
Thanks for the links.  I read them, and they're interesting.

I'm most curious, though, about old-school unregulated supplies of the type you'd find in vintage tube gear-- namely, what sorts of things give the best results in those locations (both things that can be done to upgrade/update older tube gear, and things that we're limited by physical layout).

When it comes to tube amplifiers, here's what I know-- I've always heard it recommended that there be two separate star grounds-- one for the power amp and one for the preamp.  I've heard that the center tap of the B+ winding of the transformer and the main first filter capacitor should be grounded in the same physical place when possible, and that one can get good results by terminating all preamp grounds at the input jack.  I've also heard that at the very least the cathode of the first preamp tube should be grounded at or very near the input jack. I think the idea is to keep high current grounds away from low-level grounds, but my understanding is limited.  I've also heard that the main safety ground (3rd wire on the AC line cord) should be connected to the chassis right at the point where it enters the amp, using as short a wire as possible.

I wonder the following sorts of things, specifically.  First is this-- while I've stated my understanding of the 'rule' for center-tapped B+ windings (and tube rectifiers) above, what if you have a solid-state bridge rectifier?  Where should that be grounded... also at the first filter cap?

What about something like an unregulated tube mic power supply, where you might have no power amp but rather a DC heater supply and a higher-voltage supply to feed the plates?  How should this all be grounded, both within the supply itself and with respect to feeding common along the cable to the mic itself?  Should the slightly higher current (but still relatively low current) heater supply grounds be kept as separate as possible from the B+ and capsule polarization voltages?  Assuming separate grounds, In the microphone itself, should one, the other, or both attach to the microphone case, for lowest noise?  I assume the audio ground (pin 1 XLR) should connect to the main high voltage ground?  Or should it have its own chassis connection point away from all power supply grounds?

etc. etc.  These are the sorts of things I've never understood.  If someone had the knowledge/time/inclination to correct any of my misconceptions or add to my knowledge, I'd be very interested.
 
I would suggest that designers never think of ground traces or wires as a single voltage node, but as a distributed resistance. So we must account for all currents flowing in ground traces or wires, that will generate voltage drops due to ohms law.

I have zero experience designing tube circuitry, but in general design, I just accept that voltage drops will happen in ground runs that are carrying current, and selectively pick where reference points are picked off. Brute force design is to make wires low resistance to keep drops small. Design finesse is to design such that no current is flowing in reference traces.

While I don't expect designers to visualize each electron/hole, it is perhaps useful to think of a ground system as plumbing carrying water, not some simplistic monotonic voltage node. When water is flowing in a pipe, there will be pressure drops from flow rate and pipe size.


I hope this helps. try to visualize how the PS charging currents flow through the system and wiring.

JR
 
Thanks for that.

So, a question:

Scenario A--you have a small-ish trace on a PCB.  Scenario B--you are using a thick aluminum chassis as a ground plane.

For a given distance between ground points, it's likely that more voltage drop occurs in scenario A, correct?  Or is it not that simple?  

I like the idea of designing so that no current is flowing in reference traces.  Is there any way you could scratch the surface, in a very general manner, of how one might begin to think about accomplishing this?
 
soapfoot said:
Thanks for that.

So, a question:

Scenario A--you have a small-ish trace on a PCB.  Scenario B--you are using a thick aluminum chassis as a ground plane.

For a given distance between ground points, it's likely that more voltage drop occurs in scenario A, correct?  Or is it not that simple?  

I like the idea of designing so that no current is flowing in reference traces.  Is there any way you could scratch the surface, in a very general manner, of how one might begin to think about accomplishing this?
Not to keep repeating myself, but follow the current. If there is a wire between the transformer center tap, and common point between reservoir caps, we do not expect the voltage to be the same at both ends of that wire. One or the other can be our arbitrary ground point, but never both.
----

I am not a tube guy but I expect there are similarities.  Most active circuitry involves differential or + and - inputs. When passing signals between one local ground and another, a + and - (or ground) signal reference is passed to the next differential circuit.

A unity follower has no negative input so the ground reference telescopes from the previous stage. A simple gain stage has a negative input typically where the gain resistor is connected (either supply or ground).

Note: even something as simple as a transformer is differential, having a + input that is relative to it's - input.

JR
 
There are two types of ground currents that can develop noisy voltages.
Magnetically-induced and rectifier-induced.
Both types of noise are eliminated by using hierarchical grounding, i.e. the reference point of the output of the first stage must be connected as strongly as possible (shortest) to the reference point of the input of the second stage, and so on.
Not so easy as it seems when the input and output have the same ground!
In addition, magnetic noise must be tuned out by properly orientating the mains xfmr, using low-induction xfmr, maybe humbucking.
Rectifier-induced noise is generally managed by proper thinking. The connections going to the filtering cap carry horribly distorted currents, so any of these connections is NOT equipotential. The voltage across the cap is the cleanest, that's why connections must be made as close as possible. assume that any point along the connections from rectifier to cap or from center-tap to cap is polluted.
Vacuum or solid-state makes no difference; principle is the same.
 

Attachments

  • rectifier noise.jpg
    rectifier noise.jpg
    77.6 KB · Views: 39
abbey road d enfer said:
...The voltage across the cap is the cleanest, that's why connections must be made as close as possible. assume that any point along the connections from rectifier to cap or from center-tap to cap is polluted.
epiphany moment...

I always assumed tieing everything back closest to the CT would be the best solution. My reasoning being - closer to the outlet - closer to the ground = cleaner ground. But that explanation makes more sense, and now I see why.

Grounds are one thing I'm trying to understand better too - but I think its slowly beginning to make more sense. If we think about it like a flowing water system(like JR says), then I try and think about grounds like maybe if the system backed up, trying to make sure it doesn't flow back into anything important by the size and location of the traces and how they are connected...
 
Sewage is not exactly the plumbing analogy I was going for... more like water pressure drop as relates to flow volume and pipe size, but whatever helps you visualize what is going on is all good...

While this has nothing to do with tube designs, it may help a few more light bulbs go off wrt basic power supplies. In a simple regulated power supply there are lots of nasty currents flowing between transformer and reservoir caps. Rather than try to brute force them into submission, let them be. To make my point, and this is not a suggested design, imagine a 1 ohm resistor in series with the bottom lead of your + reservoir cap, another one ohm resistor in series with top of your - reservoir cap, a 1 ohm resistor in series with the transformer center tap, and for this hypothetical design example imagine similar 1 ohm resistors in series with each  3 terminal regulator's ground legs.

Now connect all these 1 ohm resistors together, this becomes your nominal PS ground. The actual transformer center tap will be off wiggling at it's own voltage 1 ohm away, the cap resistors will have voltage across them them but so what?. The regulator ground resistors will have very little (hum) current in them, so both regulators will be pretty much ground referenced to that one point. 

Now imagine we connect another 1 ohm resistor from this single common ground node out to our circuitry as the power ground reference.  Of course we need to be alert that all current flowing in this ground reference resistor will alter the voltage between the two ends of that 1 ohm resistor.

This is in a nut shell what is going on in a well designed regulated PS. Of course the 1 ohm resistors everywhere will be milli ohms of PCB traces, or wires, but the relationships are all the same... the voltage drops in several of those connections don't matter as long as the two regulators get a clean shot to whatever is our arbitrary ground reference and no other ground currents are flowing in those particular paths. Massive brute force ground plane layouts for PS sections, while promising low impedance, and looking good does not insure that cap and transformer center tap currents, don't corrupt the path between regulator grounds and arbitrary ground node. The great "looking" layout may not work as well as some uglier simple bunch of single traces connecting the parts together in an orderly fashion steering currents where they need to go without causing trouble.

Telescoping out a power ground reference is a different matter, and I prefer to think of any product as having multiple local grounds that we reference between. Of course ground design isn't completely this simple, but it isn't mysterious, or some trial and error hunt for the magic place to connect the ground wire. In a poorly planned design, there may not be a right place to connect that wire. 

I hope this helps...



JR

 
Matador said:
This write-up covers the concepts pretty well.

http://www.aikenamps.com/StarGround.html
Good article indeed, but why does he insists on calling it "star ground", when in fact he describes the notion of hierarchical ground?
I think along the years, it has become politically correct to laud star ground. I think I'll do the same: explain hierarchical grounding and call it "star"...
[aside] Makes me think of Bill Monroe, who answered requests during his shows by playing just the next number on his set list.
Some old lady would approach him and ask him to play an obscure song, he said in the mic: "And now, for the beautiful girl on the front, I would like to dedicate xxxx (whatever she had asked)" and the band knew they'd just play the next number. Don't argue with the customer. [/aside]
Star ground is justified system-wise, essentially for safety reasons. Locally, one has to use his brains.
 
For additional reading on circuit boards and grounding, Henry Ott has some online papers, or you may be able to find his books at a collage library.

http://www.hottconsultants.com/
 
abbey road d enfer said:
Good article indeed, but why does he insists on calling it "star ground", when in fact he describes the notion of hierarchical ground?

Yes, and if you read there is no specific mention that the different hierarchies must be tied together somehow. ;)
 
Back
Top