Simple power supply grounding questions

[MisterCMRR]

Thanks guys, glad you like it - and hope you find it useful!

 

[neil.johnson]

Thanks Bill, much appreciated!

The modular synthesizer community is also often fighting noise due to very less than ideal power distribution conditions inside modular synths. Made worse by the diverse mix of analogue and digital modules often poorly designed with respect to noise.

Your papers occasionally (not enough IMHO) come up as "read this and learn", but there's only so much you can do...

Cheers
Neil

 

[Yash]

Yes, awesome info., Bill, thanks. Came across your works - probably even some of these very same slides - when I was doing some audio(phile) DIY a few years where I was solely looking at making an AC Filter box for my audio playback setup (Mac -> Audirvana -> iFi DSD DAC -> DIY Single-Ended Triode Tube Amp based on a Tubelab PCB -> Totem Mites).

Needless to say, some reading and research about this brought about a whole new set of questions to go with it and soon ballooned into 'Grounding' and other considerations like noise on the chassis and what not, which I wasn't planning to do at all. There is a thread over at audiophilestyle about these explorations ("AC Filter, Grounding...") and where your content was mentioned, together with that of Neil Muncy mentioned above, Giddings, etc...

I've mostly been lurking here. I really like the new Web UI!

 

[Yash]

The modular synthesizer community is also often fighting noise due to very less than ideal power distribution conditions inside modular synths. Made worse by the diverse mix of analogue and digital modules often poorly designed with respect to noise.

Totally! Since there's no real standard, it's a total mish-mash.

I built my own Analogue Modular, including the whole case, the Linear Regulated PSU, a custom and better designed Power Distribution (copper bus-bars a la Hinton and where I eschewed ribbon connections and use Twisted Triplets instead) and home-made PCBs, except for a Fonitronik Thomas-Henry VCO-555 which I bought.

Digital Modules are therefore powered separately, and also with Linear Regulated PSUs or batteries (like the cool little Korg SQ-1).

Physically and connection-wise for patching, it is Euro-compatible, so if I want to add a commercial module I can do that if I do a little power adapter - easy little work.

 

[Winston OBoogie]

A lot of what's detailed in "Grounding and Shielding Techniques" by Ralph Morrison might be overkill for pure analogue design as is mostly talked about on this here forum but, IMHO, it's a good reference nevertheless.

 

[Bo Deadly]

AES48 defined the industry standard for shield connections on XLR pins...you can find a draft edition here: https://www.aes.org/standards/comments/drafts/aes48-xxxx-190121-cfc.pdf

(attached for posterity)

Interesting that this doc shows the 0V reference to the chassis and not the filter cap output of the PS:


See how Signal reference (REF) is connected directly to Star point (SP) and not the PSU?

This is how I've always thought it should be done:

 

[abbey road d enfer]

Interesting that this doc shows the 0V reference to the chassis and not the filter cap output of the PS:

At least it's not the dreaded misunderstood internal star distribution of 0V.

 

[MisterCMRR]

A lot of what's detailed in "Grounding and Shielding Techniques" by Ralph Morrison might be overkill for pure analogue design as is mostly talked about on this here forum but, IMHO, it's a good reference nevertheless.

Ralph Morrison is one of my heroes! My bookshelf has quite a number of his books (one of them autographed!). He not only understands the difficulty of reaching for 120 dB dynamic ranges (low-level instrumentation systems face the same problems) but he's a clear thinker and excellent writer. He's one of the giants on whose shoulders I stand!

 

[neil.johnson]

This is how I've always thought it should be done:

View attachment 80810

I suspect there are at least two problems with this arrangement:

1. You have introduced an unnecessary metal-metal junction, which can insert noise into the signal ref line.
2. All connections should be considered unreliable, so this approach is less reliable than the recommended approach.

And now you have two "SP"s, so you now have no Star Point! There can only be one....

The connection PSU->SE is also the safety connection, so if/when the PSU fails and connects its chassis to mains then how that current gets to EG (towards SE or towards the signal circuitry) will depend on the resistances of the two paths - the lower resistance path will receive more current. You really don't want the path through the signal circuits to be the lower resistance, so do you insert additional resistance into it to protect the circuits? If you look at the recommended circuit the fault current from the PSU chassis goes direct to the SP and then to the EG. The path up to the signal circuits doesn't get a look-in.

 

[neil.johnson]

Ficchi, Ott, Morrison, Motchenbacher... all good background, then on to AES48 and Bill's slides. With all this incredible information out there it amazes me we still have "grounding" issues (which is just a branch of signal integrity).

Neil

 

[MisterCMRR]

Interesting that this doc shows the 0V reference to the chassis and not the filter cap output of the PS:

View attachment 80808
See how Signal reference (REF) is connected directly to Star point (SP) and not the PSU?

This is how I've always thought it should be done:

View attachment 80810

Always think about where the currents, especially the noisy ones, flow. Lots of noise voltage exists on "line" and "neutral" with respect to "safety ground," and this causes a noisy current to flow through the PSU to its DC outputs. This current should be given an intentional path with the lowest possible impedance back to safety ground (where it must eventually return to the power line). If you let this current flow through the path from "SP" to chassis/safety ground, the voltage drop across that path will now become noise between signal-reference and chassis - adding additional small noise voltages to the signal reference network on a PCB (or other paths) as it finds its way to safety ground elsewhere.

Another often forgotten fact is that noise voltage on the common terminal of the PSU is added (via the bypass caps at the PSU outputs) to all its DC output rails. These connections, via the local supply-rail bypass capacitors on typical PCB designs, causes this noise to be coupled into the PCB signal reference (ground) network. The impedances are very small here (milli-ohms) and noise currents small (a few mA peak typically), but only a few micro-volts between points on a signal reference (ground) can spoil noise performance when you're looking 100 dB down at the noise floor!

I used an interesting strategy to avoid this kind of "reference ground contamination" when I designed the Desper Pro Spatializer back in 1993. The power rails on each system PCB (module) were locally regulated on each PCB (using a TL431 shunt regulator - basically a programmable, extremely low dynamic-impedance zener diode). It was fed by a single transistor current-regulator, set to supply enough current for the circuitry on the PCB plus 10 or 15 mA for the shunt regulator. The current source was a very high impedance path for any AC noise on the main power rails (which were set a few volts higher than the PCB rails to give the current sources some "headroom." This scheme completely decoupled each board from the main PSU and kept all the noise that would otherwise be dumped onto the main system reference ground, adding system noise. No "experimentation" was ever necessary to keep the device's output noise at the theoretical limit. And bear in mind that this system had a main board bristling with digital processing. The signal path in the Spatializer was entirely analog - it used complex digitally-controlled VCAs. The devices are still around - I have a used one in-house right now for repairs.

Bottom line: Avoid noise headaches, make the chassis the star point in a signal processing box!

 

[MisterCMRR]

Ficchi, Ott, Morrison, Motchenbacher... all good background, then on to AES48 and Bill's slides. With all this incredible information out there it amazes me we still have "grounding" issues (which is just a branch of signal integrity).

Neil

Right on Neil! But you can drag a horse to water, but you can't make him drink!
What really irritates me sometimes is that Europeans tend to dismiss "star grounding" techniques altogether. I believe it's due to the strong influence of the telecom/IT/RF design community (Armstrong and others) who champion a "mesh" grounding technique - which works fine in high-frequency systems but they believe it applies to ALL systems! What's unique to audio systems is that mains frequency and it's significant harmonics fall into the audio band. IT and higher frequency systems can afford to ignore or tolerate it ... we can't!!

 

[abbey road d enfer]

What really irritates me sometimes is that Europeans tend to dismiss "star grounding" techniques altogether.

I'm not so sure about it. If you watch my sig, I'm not dissing star ground, but star ground is often understood and presented as a magic means of producing the best signal integrity.
I've seen examples of designs where a number of PCB's were star-grounded, which resulted in significant (AC) voltage differences between grounds, that actually end up being added to the signal. That was in complete violation of "signal follows ground".
I agree that star ground is a must for mains, but must be taken with a large pinch of salt when it comes to audio or video signals inside equipment.

I believe it's due to the strong influence of the telecom/IT/RF design community (Armstrong and others) who champion a "mesh" grounding technique - which works fine in high-frequency systems but they believe it applies to ALL systems!

Counterweight to those who champion star grounding and believe it "applies to ALL systems!"

 

[MisterCMRR]

I'm not so sure about it. If you watch my sig, I'm not dissing star ground, but star ground is often understood and presented as a magic means of producing the best signal integrity.
I've seen examples of designs where a number of PCB's were star-grounded, which resulted in significant (AC) voltage differences between grounds, that actually end up being added to the signal. That was in complete violation of "signal follows ground".
I agree that star ground is a must for mains, but must be taken with a large pinch of salt when it comes to audio or video signals inside equipment.

Counterweight to those who champion star grounding and believe it "applies to ALL systems!"

I'd assert that the problem you cited with a number of PCBs was likely a prime example of ignoring the need to de-couple power supply rails at the PCBs. But this gets into the nuances and details of design. Common-impedance coupling is the dominant problem in low-frequency systems (the average wire stays "resistive" up to at least 5 kHz) and single-point ground referencing can completely eliminate it. Above 5 kHz, wires look "inductive" (higher impedance) and "grounding everything everywhere" (or "mesh") is the only way, short of a huge "ground plane" to control magnetic and electric field coupling. There's a range of appropriate frequencies for each technique. One reason I've championed the idea of "hybrid" grounding - it was our work in the AES standards committee that prompted Neutrik to design the "EMC" XLR connector, which mitigates RFI issues when strict one-end-only rules are applied to XLR cables (common-mode conversion results when balanced cable shields are tied at both ends, but this is a purist consideration ... see my 1995 AES paper for full details - I don't want to start an argument here!!

 

[JohnRoberts]

There have been several good text books written on the subject of grounding (and shielding). I have a couple still on my bookshelf from the 70s.

A simple grounding question is an oxymoron....

JR

 

[Bo Deadly]

We advise people almost daily about proper grounding in audio devices and stress that it is, without doubt, the most important hurdle that a new builder must overcome. There's really no point in building something if it's noisy for whatever reason. I myself am obsessed with minimizing noise. So if my understanding is incorrect I'd really like to fully understand why so that I'm not giving people bad advice.

Always think about where the currents, especially the noisy ones, flow. Lots of noise voltage exists on "line" and "neutral" with respect to "safety ground," and this causes a noisy current to flow through the PSU to its DC outputs.

Can you explain the origins of this noise more precisely? Is that common mode noise making it through the transformer or capacitive coupled or noise of the rectifiers?

This current should be given an intentional path with the lowest possible impedance back to safety ground (where it must eventually return to the power line). If you let this current flow through the path from "SP" to chassis/safety ground, the voltage drop across that path will now become noise between signal-reference and chassis - adding additional small noise voltages to the signal reference network on a PCB (or other paths) as it finds its way to safety ground elsewhere.

If REF (what we usually call "0V" around here) is the downstream end of the power supply output ground, then it's not crystal clear to me why we should be concerned about noise relative to the safety ground. Like you said, this noise has to make it all the way back to the power line. So attaching REF to the chassis over a few inches of a much longer wire seems insignificant to me. Otherwise does the performance of the device then not depend almost entirely on the rejection of noise at inputs and outputs (like CMRR of input using a Jensen transformer of course!)?

Or is the noise issue created by connecting REF to the PS output as I proposed exclusively regarding phantom powered devices because phantom power returns through the chassis (SE) and therefore has to go through the PS output to safety ground connection which has the superimposed said mains noise?

Another often forgotten fact is that noise voltage on the common terminal of the PSU is added (via the bypass caps at the PSU outputs) to all its DC output rails.

What is this "noise voltage" relative to? If the common terminal of the PSU is connected to REF as I have proposed then would the noise voltage not be relative to itself which is to say there would be no noise voltage at all. Please enlighten.

One rule about grounding and more generally about minimizing noise that is used judiciously here is the "return follows supply" rule which states that return traces / wires should be as physically close / symmetrical to their supply lines. It seems like the scheme used in that AES doc would violate this rule as illustrated by the below graphic where the red area represents a larger than necessary loop area:




Why is this not a concern?

using a TL431 shunt regulator - basically a programmable, extremely low dynamic-impedance zener diode).

That's going to by my tea table reading thanks.

Finally, I have to ask if any of your comments would be different if the supply were an SMPS? It has been my experience, and the experience of many others here, that SMPS (a good one anyway) is superior to a linear supply in every category (with the exception of EMI if you try to put too close to sensitive bits). I suspect the answer is no as an SMPS is ultimately the same components albeit in a different sequence but SMPS are used almost exclusively in new designs so it would be important if your advice where different in any way.

 

[Yash]

It is my experience that for audio, especially in the unbalanced setups, a simple Linear Regulated PSU usually pollutes much less than almost any commonly available SMPS.

I abhor the latter for audio. They pollute widely, in both directions.

 

[Bo Deadly]

It is my experience that for audio, especially in the unbalanced setups, a simple Linear Regulated PSU usually pollutes much less than almost any commonly available SMPS.

I abhor the latter for audio. They pollute widely, in both directions.

That is completely false. If you get the right SMPS and you use it properly, it will definitely out perform a linear supply in every way. No question. But again, there are rules and if you don't follow them, you can easily get noise:

Rules for Using SMPS for Audio

 

[Sc]

Learnt a lot again about grounding - thank you. Unfortunately I do not own any of the original german studio / broadcast cassettes of the late tube technology phase like V72 etc. with the Tuchel connectors. But it seems to me, that most of the schematics have zero Volt lines not connected to the earth/case ground. Maybe I'm wrong, I've never seen the racks either, but if not, what was the reason for this seemingly strict separation of Zero and case/earth ?
edited: case not chassis - chassis might have been on 0V.

 

[Matt Syson]

Hi
As an 'European' I dismiss 'star grounding' to some extent in that unless it is used intelligently it cannot solve noise issues in a complete system. In a usual 3 dimensional world and outside the realms of superconductivity, everything has resistance and inductance and capacitance to all other conductors so you have one heck of a network of interconnected parts so to get an 'optimum' low noise, low distortion 'system' you must consider where all currents will flow and electrostatic fields of course, from essentially DC to light, and beyond. You have to decide WHICH electron will be your reference point.
I remember in a discussion of a 'new' audio mixer 40 years ago the project the chief designer making calculations on the possible effects of the 'noise' current flowing from multiple LM7815 series regulators to 'ground' (the current that flows from the regulators 'ground' or common terminal) in an attempt to decide whether local on board regulation would be advantageous over a 'simpler' well regulated single supply.
As a designer you have to consider where all currents SHOULD flow and inductive and capacitive interactions with all other conductors. As they are all aerials either transmitting or receiving the impact of this has to be considered too and measures to limit or at least control the effects.
Separating supply 'zero' and chassis/case would be done to limit intrusion of 'noise' into supply lines such that if there were potential differences between various pieces of the chassis (quite likely) they would not be impressed onto the DC power system to a fair extent. Bearing in mind that there will be 'leakage' of capacitively coupled 'noise' from various sources anyway. When testing mixers in the past the first check is that the chassis is not connected at all (DC wise at least) to any power rail, analogue ground, digital ground or any other 'grounding paths (summing mix analogue ground).
Matt S