Grounding cable shields

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

JMan

Well-known member
GDIY Supporter
Joined
Oct 26, 2019
Messages
715
Location
Midwest, US
Hi Everyone,

I have a question regarding the grounding of shields. This is both practical for a build I'm thinking about, and theoretical (if there is any general theory to be gleaned here).

Below, I've attached a simple diagram of how the various ground points of the project are intended to be connected. However, in an effort to "overbuild" somewhat,* I have chosen to use shielded cable for the various front panel connections. There are not enough convenient ground pads on the main boards to connect the shields on that end (they are probably overkill and not completely necessary, and thus were not anticipated on the pcb), so my thought was to connect them all to a bus wire at the front panel instead end and run that off to ground. And yes, they will only be connected at that end. My question is this: as you can see, several of the other ground points "touch base" at the PSU, from which they are all connected to the bolt on the chassis. Does it make sense for me to also send the ground bus to that same point on the PSU, or directly to the chassis? Does it make a difference / is there a reason why, either way?

*The primary reason for the shielded wires is that my front panel layout is a little different and many of these cables will run near-ish to the AC lines of the power switch, and I'd like some extra insurance against that noise.
 

Attachments

  • Grounding Scheme.png
    Grounding Scheme.png
    266 KB · Views: 115
Thats fine. Ideally the shield pin of each connector (but not a signal ground) should be connected to the chassis over the shortest possible wire. This is because RF can be radiated within the enclosure and the wavelength of RF is determined by the length of the wire. The shorter the the wire, the shorter the wavelength, the higher the frequency, the more limited the RF noise. The best way to do this is to use XLR connectors that have a metal spike in the screw hole that stabs the metal panel. But you sort of have to plan for that in advance.

But otherwise yes, the shields are an extension of the chassis and therefore you want to connect them to the chassis bolt and not the power supply.

However, these connections use 48V, that returns through these wires. So you want to use a beefy wire. If the bus wire is smallish, you might even twist two strands together to double up. And the wire from the chassis bolt to the PS ground should be fat. Like 18 AWG or lower.

Also not that, even though you're not showing any controls, sometimes a shield can be used as signal ground in which case none of the above applies.
 
Hey Bo, thanks for your reply. I probably rambled too much in my question and buried the lede. I can certainly attach the xlr pin 1 points closer. Also, this isn’t a preamp, so no 48v to worry about.

My actual question is about the shielded connections from the main board to the front panel controls. Those shields are NOT connected to the main board, so I’m planning to bus them together at the front panel. From there, does it make more sense to run them directly to chassis, or through that PSU ground point?
 
My actual question is about the shielded connections from the main board to the front panel controls. Those shields are NOT connected to the main board, so I’m planning to bus them together at the front panel. From there, does it make more sense to run them directly to chassis, or through that PSU ground point?

I understood your question. The answer is the chassis. Just like you have drawn up.

But again, and I'll elaborate more deeply, the IDEAL arrangement would be to have each pin 1 connected to the chassis right at the entry of that socket. Meaning these wires would be like 2cm long (or use the XLR with integrated spike and then you don't need to wire anything). Then the shielded cables from the PCB should only be connected to 0V at the PCB and not at the sockets. Otherwise the shields could actually radiate RF inside the enclosure. In fact, if the sockets are line level signals and not mic level signals, you might be better off not using shielded cables at all.
 
Hey Bo,

I guess I’m not understanding correctly. I DO understand what you’re saying about the input/output xlrs and pin 1, and I can make that change, but that is not my question. (And just to be absolutely, explicitly clear, I am not asking about the xlr-to-pcb connections).

I’m not sure what your are referring to when you say “sockets,” unless this is again related to the xlrs.

Also, you said “The answer is chassis. Just like you have drawn up.” But I specifically didn’t draw the ground connection for the shields of the wires that run to front panel controls, since that was where I was uncertain and asking about. Maybe I am missing something obvious...

So I will try once more, in case it is still me not understanding or articulating myself well:

I am asking about using shielded wires to connect MAIN PCB to FRONT panel controls in an effort to reduce noise. Those shields do NOT connect to ground pads on PCB. They DO connect to a common bus wire nearer to the front panel. My question is where should that bus wire then lead — PSU, or direct to chassis, or elsewhere, or makes no difference?
 
I am asking about using shielded wires to connect MAIN PCB to FRONT panel controls in an effort to reduce noise. Those shields do NOT connect to ground pads on PCB. They DO connect to a common bus wire nearer to the front panel. My question is where should that bus wire then lead — PSU, or direct to chassis, or elsewhere, or makes no difference?

Ah. Ok. I missed the "front panel connections" part.

So in that case, assuming that these shields are not tied to shields that run outside the enclosure (like pin 1 of XLR sockets), technically I think it would be better (given your particular scenario) to tie your shield bus to the power supply ground and NOT the chassis bolt. This is because the wire connecting the chassis bolt to the PS ground might have high frequency noise from the outside across it. So you don't want to distribute that through the shields of your controls. All of the control signals are relative to PS ground so you want your shields to also be relative to PS ground.

But this again assumes that the shields are really just shields and they're not actually connected to 0V (aka signal ground) through any other path other than at the PS ground (given your particular scenario). Generally you want to avoid making loops. Especially involving 0V. Your shields should be branches of a tree. So don't connect even one of the shields also to the PCB for example (given your particular scenario).

However, this is still the best alternative given your particular exceptional scenario where you do not have PCB ground points for shields. Using the local 0V for the shield is the correct way to actually shield those panel control wires.
 
My question is this: as you can see, several of the other ground points "touch base" at the PSU, from which they are all connected to the bolt on the chassis.
It depends what point of the PSU. The connection should be to a clean node, which is usually the reference point of the regulators or the last caps in the chain. The junction of rectifiers or xfmr secondary must be avoided.
Does it make sense for me to also send the ground bus to that same point on the PSU, or directly to the chassis? Does it make a difference / is there a reason why, either way?
It shouldn't make a difference if the star point is well chosen and no current is allowed to circulate.
 
Thanks, Abbey!

I believe that both the PSU ground point and the star point fit the description of “well chosen.” PSU is indeed a clean node, and the star point is the typical spot right next to the IEC. But of course, “well chosen” always seems to involve considerations that you more experienced/knowledgeable folks are aware of that I wouldn’t even know to take into account, and thus I end up asking these questions! :unsure:🤪

I believe what I’m going to do is send those shields to the PSU point to start, per Bo’s advice above. If I have any problems, I can always experiment with moving them elsewhere.
 
The primary reason for the shielded wires is that my front panel layout is a little different and many of these cables will run near-ish to the AC lines of the power switch, and I'd like some extra insurance against that noise.

The primary pickup mechanism for power supply noise is magnetic induction, not capacitive, so shielding those control wires will provide no benefit. What will help is making sure the power wiring is tightly twisted, and the control wiring is tightly twisted, to reduce inductive pickup.
 
It depends what point of the PSU. The connection should be to a clean node, which is usually the reference point of the regulators or the last caps in the chain. The junction of rectifiers or xfmr secondary must be avoided.

It shouldn't make a difference if the star point is well chosen and no current is allowed to circulate.
I am working on a similar project; a rack with two NTP 170-140 compressor/limiter modules, powered by a Kniel CD 15.0,5 linear PSU. Both my NOS 1.04.06.242 and Kniel´s modern version 120-003-02 are dual +/-15V, hence seemingly lacking a dedicated 0V...?
You say the 0V should ideally be a clean node. I have connected the PSU´s +O/P2 and -O/P1
using a synthetic(?) 0V, ref. picture. This 0V goes to the NTPs, and is further connected to the
star point screw, where immediately meeting the AC inlet PE and chassis/all XLR cable screens.
The Kniel PSU works fine, yet somehow this 0V solution feels kind of bodged - is there a better and/or safer 0V solution using this dual Kniel PSU?
 

Attachments

  • KNIEL CD 15.0,5; 120-003-02 & 1.04.06.242, bipolar connection.jpg
    KNIEL CD 15.0,5; 120-003-02 & 1.04.06.242, bipolar connection.jpg
    105.5 KB · Views: 80
The primary pickup mechanism for power supply noise is magnetic induction, not capacitive, so shielding those control wires will provide no benefit. What will help is making sure the power wiring is tightly twisted, and the control wiring is tightly twisted, to reduce inductive pickup.
Interesting! Okay, so now I’m wondering, why do I see projects that use shielded cable internally? I mean, from what you said, it must be because there is some capacitive source of noise in those cases. So how would I recognize when one or the other is likely to present itself? (I’m sure this is a terribly broad question, but like, are there typical examples?)
 
This is because RF can be radiated within the enclosure and the wavelength of RF is determined by the length of the wire. The shorter the the wire, the shorter the wavelength, the higher the frequency, the more limited the RF noise.
Could you please explain this a bit more as the length of a wire does not appear in the standard wavelength formula?

However, just o add to your comment on connecting all the grounds at a single chassis point is that using multiple ground points in a chassis means varying many circulating currents which is an invitation to trouble.
 
Could you please explain this a bit more as the length of a wire does not appear in the standard wavelength formula?

The shield wire is like a transmitter. So it must be connected to the metal enclosure as close as possible to the point of entry. If the wire is a piece of bus wire that runs all the way from the XLR socket over to the chassis bolt, that is not ideal because if it is say 30cm, then that would appear to be in the upper UHF range:

radio-frequency-bands.png


A simple f = C/λ sort of calculation with 0.3m would yield 1GHz which aligns perfectly with the above chart.

However, just o add to your comment on connecting all the grounds at a single chassis point is that using multiple ground points in a chassis means varying many circulating currents which is an invitation to trouble.

Is this a question? Right. If you have a loop of wire, then magnetic fields will induce currents like a transformer winding so for signal wires you definitely want to minimize loops. There are always lots of tiny loops in ground and supply lines but if they're usually pretty small. Although when I do a ground plane layout and I find that the ground plane has a big hole in it such that it just makes a loop around the edge of the board, I will deliberately make a break in the plane somewhere to break the loop. Not sure if that's overkill.
 
Last edited:
Bo,

You have to be sure about the formula as it is the formula used for wavelength. The chart you gave is not about the actual value, it shows more like the range.

Your comment is problematic because the wavelength has no dependency on wire (conductor) length. A 1MHz RF will appear on a given length of wire in the same way a 100kHz will. So, I humbly recommend that you revise your comment on this.

But the inductance and capacitance of a wire (conductor) is another story as these will have effect on the signal, but not in the sense of reducing or increasing the wavelength, but as effect on signal attenuation.

However, in terms of XLR, ideally pin1 should be connected to the chassis immediately at the point where the XLR is mounted. If this is not possible (perhaps plastic enclosure-bad idea) then yes, keeping the chassis earth wire as short as possible would be a good practice. But inside a metal case, which is already grounded, I do not think one needs to be too fussy about the length (within reason).

My other comment was not a question. It was to complement your comment.
 
However, just o add to your comment on connecting all the grounds at a single chassis point is that using multiple ground points in a chassis means varying many circulating currents which is an invitation to trouble.
A big distinction must be made between shields that are only electrostatic shields, that can have one or teh other or both extremities "grounded", and those that are also used for return current continuity. When submitted to EMI/RFI, the former are usually less problematic than the latter, where induced voltages are directly added to the signal.
 
Your comment is problematic because the wavelength has no dependency on wire (conductor) length

I believe "Bo Deadly" was just being too brief. What he is trying to convey is that a wire extending into the enclosure becomes an antenna, and the frequency at which the antenna becomes a reasonably efficient antenna depends on the length of that wire. If the interfering frequencies present are frequencies where the antenna does not radiate efficiently, then there is not much problem. If there happens to be noise current flowing at a frequency where the wire becomes an efficient antenna, then that wire will re-radiate the noise into the interior of the enclosure.
 

Interesting! Okay, so now I’m wondering, why do I see projects that use shielded cable internally? I mean, from what you said, it must be because there is some capacitive source of noise in those cases. So how would I recognize when one or the other is likely to present itself? (I’m sure this is a terribly broad question, but like, are there typical examples?)
Capacitive coupling dominates at higher frequencies. "Switching power supplies" generally have boatloads of high frequency noise generated from the switching transients. Digital circuits too. Usually, the faster the switching (or clock,) the worse the problem gets.

If your supply is a standard transformer - rectifier - linear regulator type, probably not much at the higher frequencies to worry about. You'll get a bunch of harmonics at multiples of 60Hz from the nonlinearity of the rectifier diodes, but it usually dies off above maybe the 'low-midrange.'

If you're using a switching DC/DC converter, shield it... Shield anything anywhere near it. Putting it in a different box helps too. Those things can easily sing up into the MHz range (or higher!) Twisting nearby pairs can actually help reject E field too, but you'll need some common mode rejection like from a transformer or a diff-amp at the end of that line to clean up the signal.
 
@abbey road d enfer

Certainly. I have been loose there too.

@ccaudle
I understand that. But the description "wavelength of the rf depends on the wire length" is still problematic.

As I mentioned before pin 1 of the XLR should be tied to the ground immediately where it is mounted onto the chassis. The scheme JMAN has given in his diagram is already a bad idea in terms of pin 1 earth connections.

If he is using a plastic enclosure (which is even a worse idea) then he may need to do that, and yes keeping the earth wire as short as possible would be a good design practice but not hugely important if it is a few inches longer.

Also, the wire extending inside the case will not radiate as now the wire is a receiving antenna and is terminated at the chassis/earth point (at the IEC). So, the rf noise currents will run towards chassis earth point.

As for the screens for the signal wires used inside the case the situation is slightly different. Good practice would be to tie the screens on one end, and at a single point at the chassis/earth point near the IEC, so that all noise currents would run towards that point. The mistake would be to earth them on varying chassis points within the case. Now you have varying ground paths and that is an invitation to a problem.
 
Back
Top