cable sheild as signal ground. Internal, short runs

So

I think I understand this now. Ideally, we never want the shield to be connected to 0V signal ground, right? (except at the one star point)? All shields should be connected to chassis ground, and only on one side?

What about short runs internally, like from a pcb board to the front panel or something?

I ask because I'm realizing there are several common pieces of equipment that i have where I connected the shield of a wire (say to front panel on G1176) where the shield actually connects back  to the board's 0Vu.

Search for Ian's (ruffrecords) grounding article it goes over most of this is excellent detail.

Screens 101.

If the signal is balanced the screen is a screen not a signal carrying conductor. You connect it to chassis where you want all the cr** to go.

(most people think a balanced signal has two signals referenced to ground. This is not true. A balanced signal exists between the hot conductor and the cold conductor only. Ground is not required.

If the signal is unbalanced the signal exists between the signal conductor and analogue 0V. If you want to take that signal somewhere else inside the system (where the analogue 0V of the receiver is already connected to the same analogue 0V as the sender) then you only need to send the signal.

BUT this is very risky because there is a good chance the two analogue 0Vs are not at the same potential due to the resistance between them. The unbalanced receiver amplifies the signal between the signal conductor and its own analogue 0V. So this potential difference between the two analogue 0Vs will be added to the signal and appear as cr** in the receiver. This is the regular hum loop problem.  It does not matter which end you connect the screen, the loop still happens. It is probably best to connect the screen to analogue 0V at both ends because it reduces the resistance between the two analogue 0Vs.

There are several approaches to solving this problem.

Old days. Keep the resistance of the ground as low as possible. Do this by using fat copper or brass bus bars for analogue 0V. Minimise the number of unbalanced signals. We did both of these at Neve when I was there back in the 70s. Channel amps were transformer balanced in and out. Routing modules were transformer balanced in. The only unbalanced stuff was from routing modules to the buses and then to the bus amps.

Modern approach. Still keep analogue  0V resistance as low as possible but also use differential receivers and send both signal and local analogue 0V from the sender to the receiver. This way the receiver looks at the difference between the signal and the sender's analogue 0V so it ignores the potential difference between the two 0Vs. In this system you connect the screen to analogue 0V at the sender and to cold input of the differential receiver.

None of this fixes the problem for buses, especially in a large consoles simply because there are multiple senders to each bus so the only thing to do is keep the resistance of analogue 0V as low as possible.

Cheers

Ian

Thanks Ian for all that info. That's really great. And I HAVE read your very helpful grounding article.

From that post I can't tell if you're talking in general about the the full system (mixer, gear etc. ) or if you were talking about internally in a single piece of gear. Or both.

My inquiry has to do with the inside of one piece of equipment. So ostensibly the same 0VU reference (?)

Let's say we have a common +/-16VDC supply for an API unit or something. When connecting the power supply to a card, you'd want to connect 0V to 0V, but is it okay for that connection to be carried by the shield of a common 2 conductor (+ shield) cable?

Or this might be a different scenario internally, where we have the shield connected at one end. (Not carrying a ground reference) just the shield of a short cable from a pcb board to the front panel, like in say the comp/limit switch on an LA2A. This can be just a simple spdt switch, so conveniently, a 2 conductor cable (+ shield) can be used. I noticed however, that the shield for this cable made less noise when connected to 0VU at the circuit board, than to the chassis ground at the front panel.

Just wanted to chime in and say I had the same issue and Ian as well as many others were very helpful with their suggestions.  They convinced me to lay a large 0v pour instead of chassis pour and all my long-ish unbal traces were replaced with pads and shielded P2P wire tied to 0v only on one end.

When connecting the power supply to a card, you'd want to connect 0V to 0V, but is it okay for that connection to be carried by the shield of a common 2 conductor (+ shield) cable?

Click to expand...

My intuition says make power and its ref as low a resistance as is reasonable, which is to say if you can fatten em' up with dedicated lines, that couldn't hurt. Shield should be shield, reference should be reference.

JW said:

Thanks Ian for all that info. That's really great. And I HAVE read your very helpful grounding article.

From that post I can't tell if you're talking in general about the the full system (mixer, gear etc. ) or if you were talking about internally in a single piece of gear. Or both.

Click to expand...

I was mainly talking about inside a piece of gear.

My inquiry has to do with the inside of one piece of equipment. So ostensibly the same 0VU reference (?)

Let's say we have a common +/-16VDC supply for an API unit or something. When connecting the power supply to a card, you'd want to connect 0V to 0V, but is it okay for that connection to be carried by the shield of a common 2 conductor (+ shield) cable?
/quote]
In principle it does not matter what you use for this connection as long as it is low resistance but I am not clear what is going down the centre conductor of this cable?
[quote
Or this might be a different scenario internally, where we have the shield connected at one end. (Not carrying a ground reference) just the shield of a short cable from a pcb board to the front panel, like in say the comp/limit switch on an LA2A. This can be just a simple spdt switch, so conveniently, a 2 conductor cable (+ shield) can be used. I noticed however, that the shield for this cable made less noise when connected to 0VU at the circuit board, than to the chassis ground at the front panel.

Click to expand...

Specific cases are very hard to comment on. A lot depends on how the chassis cos connected to 0V (assuming it is). The LA2A comp/lim switch is essentially an unbalanced internal signal so any screening of it should be connected to local 0V.

Cheers

Ian

JW said:

So ostensibly the same 0VU reference (?)

Click to expand...

VU does not mean volts. V means volts. VU means volume units which has to do with signal levels.

JW said:

Let's say we have a common +/-16VDC supply for an API unit or something. When connecting the power supply to a card, you'd want to connect 0V to 0V, but is it okay for that connection to be carried by the shield of a common 2 conductor (+ shield) cable?

Click to expand...

Ideally the chassis ground should only connect to the power supply ground at one point. Usually earth ground coming in from mains is connected to the chassis right next to the mains plug. Then a wire from there to power supply board is the one point where they connect. Then from the PS to the main board.

Shields should only be connected at one end. If a shield is connected at both ends such that you are effectively connecting the chassis and PS ground together at more than one point, that creates a loop. A loop in the ground is potentially bad because a loop of wire is sensitive to electromagnetic interference.

Having said that, old guitar amps almost always have multiple ground loops formed by all of the jacks and pots bolted to the chassis but then also connected to signal ground at the inputs, effect send / return, reverb tank cables and so on.

If you want to shield a signal cable running out to the front panel, you should connect the shield at one end but not at the other. It doesn't really matter which end but my feeling is that usually the PCB end is used.

This standard practice in Urei gear like the 1176 and LA3A. You can even see how the shielding is done from looking at their schems. They're drawn like this:



You can see one end of the shield but the other end is dismissed as a squiggly line. That means the other end isn't connected. You might think the "INPUT" pot of the 1176 is an exception to this because it is connected at both ends but if you look carefully, it's actually not connected to ground at the pot end. It's completely isolated through a cap.

Here is a pic I found of the corresponding control in the 1176:



Note that in this case, because the pot also needs signal ground, it just takes it from one of the shields. So it's doing "double duty".

Now in practice, it's probably not really a problem if you connect the shield at both ends. If the ground is the same ground (meaning chassis -> shield -> chassis) that's actually perfectly ok (the chassis is where we want to shunt EMI anyway). If you have a shield connected to the PCB ground at one and and then also to chassis ground at the other end, that could pick up some hum from a transformer or some other source of EMI. But my guess would be that it's actually fine and it's not a problem. If you really want to know, you have to measure things. I measure the cr** out of everything. My favorite thing to do right now is to just make a recording and then do FFT plots with software.

But if it's a simple matter of opening it up and cutting the shield connected to the chassis, that might be a simple fix.

Okay,

Let me thank you guys for the wealth of info. And apologize for sort of being all over the place with my question asking. (Also, yeah, typo on "VU" I just meant 0V. )

Backing up a bit, it was actually Ian's very useful and clear grounding article that sent me down this bunny hole. Also one of the RANE documents. I had originally set up my studio to have the shields lifted on the inputs to all line level gear. I kept hearing it said, however, that the shields should really be connected at both ends if one can ensure that all gear is balanced and has the shield connected directly to chassis right at the inputs and outputs. I have the ability to ensure that as all my gear is balanced (outside of a few effects but those can be dealt with) So, you know, why not?

But then me realizes a few other things that I've done without thinking. Such as, previously mentioned, using a 2 conductor (+shield) cable between the +/-16VDC output of API power supply and the API board. +lead -lead, and ground, but the ground in this instance is forced to be on the shield. Is this okay? (also should be noted that this is THE point that is connected to chassis)

Also, I realized, that a couple other pieces of gear I have, like 2 ch 1176 and 4 channel Tridentish 80B eq's have daisy chained signal grounds. That is, once again, I'm using 2 conductor (+ shield) cable from the power supply (0VDC+/-16V out) to one board, and then another and then another, daisy chained. But the ground is carried always by the shield. Is there a general best practice for multi channel units in regard to the 0V power supply output? Daisy chain vs. 'homerun' grounds for each board?

I should be clear that I understand the concept that there should be one point only where signal ground connects to chassis. And in all these instances, I have my power supply output 0V (signal ground) connected shortly to chassis at the same IEC mains earth ground connector.

edit: Squarwave. Thanks! That's exactly what I was thinking about in regards to 1176 output pot. Something like that where there's dual duty with the shield. But very short connection.

JW said:

I kept hearing it said, however, that the shields should really be connected at both ends if one can ensure that all gear is balanced and has the shield connected directly to chassis right at the inputs and outputs. I have the ability to ensure that as all my gear is balanced (outside of a few effects but those can be dealt with) So, you know, why not?

Click to expand...

Right. All of the chassis grounds are connected together. But that's separate from signal ground inside each piece of gear. Within each piece of gear, you only want to connect the chassis ground to the signal ground in one spot. Otherwise, it makes a loop that can pickup noise.

JW said:

Ideally, we never want the shield to be connected to 0V signal ground, right?

Click to expand...

That constraint applies to cable runs between pieces of gear, because the shield acts as an antenna for all the EMI/RFI.
Inside a chassis, that is a completely different situation; it is perfectly legitimate to use a shielded cable between two PCB's and use the shield for ensuring the return path.
It is in fact actually better than running the return path via a star-ground 0V reference. Using both the inner conductor and shield ensures whatever interference is cancelled out by making the return path equal to the signal path.