Back to basics - Where to terminate shield 2

[123MB]

Hi All

When it comes to first principles, a cable shield is intended to carry noise currents away from places they shouldn't be, and allow them to return to their source. Would you agree with this?

In principal then, is it correct that a shield should be connected to the return terminal of the source of supply of the noise current?

Take, for example, a stand alone ungrounded DC solar system where the cables from the PV panels to the charger, and from the charger to the battery, are generating radiated noise which can inductively and capacitively couple onto nearby sensitive circuits (i.e. mV measurement)

If one was to provide shielding on sensitive cables (i.e. thermocouple) which are in the same duct as the above cables, where would you terminate the shields? If you terminated them to the chassis, would it provide a return path for the noise currents or just a high impedance path that did very little?

In the real world example, the solar charger does not have a chassis earth terminal for bonding.

 

[Skogsgurra]

There is no easy answer to your question. It all depends on what you have and what you want to achieve.

Before Power semiconductors, there were not many sources of HF pollution - except transmitters of different kinds - like radar, radio, TV and such.

Also, there were not many sensitive circuits that were affected by interference. Microphone Cables were the more common ones and a practice not to ground anywhere but at the amplifier end developed. In industrial and energy applications, the 0-20, later 4-20 mA was regarded the standard solution to all interference problems.

Later, when frequency inverters and other power inverters (like you have in your solar system) the standard became the opposite: Ground everywhere and Everything that could be grounded or bonded. That works quite well in some systems, but may be a problem in TT systems, where there are different grounds at both ends of the cable. TN systems and IT systems are less difficult to handle.

Next move was to introduce BW limitation. A thermocouple doesn't need much band-width, so simple low pass filters worked quite well. Differential amplifiers were also used, and still are, but the CMRR could be a problem since most simple differential amplifiers didn't work very well. Capacitive coupling and flying capacitors was a way towards better immunity to HF pollution and optical isolation is, if correctly done, one of the better ways.

At the same time, common-mode filters "ferrites" helped somewhat and nanocrystalline CM filters were/are very helpful up to a few MHz and are extensively used in inverter applications to avoid interference and stray currents/bearing currents that ultimately destroy bearings (EDM).

So much has been written about this and one magazine devoted to the subject is Electronic Environment where Michel Mardiguan presents basic theory and lots of case studies in the most diverse applications.

As I said - and you probably also know - there is no king's road to geometry (Euclid) or electro-magnetic compatibility. Insight, some math and experience with the system in question is what counts.


Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[MacGyverS2000]

Take two cases on a long wire connecting two pieces of equipment:
1) Shield is connected at one end only -- Radiated fields on the non-connected end find themselves on the shield... how do they get back to their source?
2) Shield connected at both ends -- ground loops are the bane of many an audio engineer.

Now, how can both be "wrong"? That discussion has kept many an engineer up at night...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[ScottyUK]

Part of the key to understanding how to combat noise on screens is to understand how it is getting coupled in the first place. Inductively coupled noise requires a loop. Radiated noise requires an ungrounded conductor, or at least one with enough impedance at the frequency of interest to be effectively ungrounded. Once you figure out what you're fighting against, it is easier to develop a solution.

 

[123MB]

Thanks all

So how about practically, for a stand alone solar ungrounded DC cabinet with very long lead thermocouple measurements being made by converters inside the cabinet, where would you connect the screens of cables?

 

[IRstuff]

Shields are supposed to be terminated at the nearest single-point ground (SPG). Often, this is simply not sufficient, and you'll need to ground it locally, as well.

We use to waffle between grounding near/far/both ends of shield, sometimes changing a few times during the course of a design. Most of our current guidelines say, ground at the SPG.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[ScottyUK]

A thermocouple is a low impedance source and temperature is a fairly slow-moving variable, so take advantage of those facts by keeping the controller input impedance relatively low and filter the hell out of it. I'd ground the shield at the controller end and leave the field end ungrounded.

 

[Compositepro]

Any current flow in the thermocouple wire over that distance would alter the signal. So I don't know that a low impedance input amplifier would be helpful. It is best to put a temperature transmitter much closer to the thermocouple junction.

 

[ScottyUK]

Compositepro - the key word being 'relatively'. Input impedance in the hundreds of k[Ω] rather than the hundreds of M[Ω]. The loss of accuracy with a source impedance of a few Ohms is less than the inaccuracy introduced by use of compensating cable.

 

[123MB]

In this example the DC negative/0v is the only return path for any kind of signal, whether it be pure DC or a noise signal. Considering this, wouldn't it be wise to connect the shields to the DC negative/0v? Or would it be even wiser to bond the DC negative/0v of the supply to the chassis/ground/bonding system?

In general, grounding the 0V is supposed to reduce common mode noise on the supply, correct?

 

[IRstuff]

As a general rule, you avoid doing that, because ANYTHING that gets onto your chassis will therefore get into your internal grounds. Moreover, your chassis is now an active part of your ground return, and shock hazards abound.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[123MB]

Now I'm definitely confused!

So for a noise path back to a DC source... your DC source has to have a chassis connection with DC blocking capacitors?

 

[Skogsgurra]

Allow me to quote myself: "There is no easy answer to your question. It all depends on what you have and what you want to achieve"

Yes, you have many reasons to be confused. There is no easy answer. And, to make things even more complicated and confusing, there was some consensus that the "Control zero" (the 0 of the then quite common +15/0/-15 V supplies) should be isolated from ground - the so called "Floating Zero". Popular in European Control systems. That didn't work out so well in the long run (static charge of the whole system or invuluntary ground connections all over the Place) so the "RC" ground was introduced. A ceramic capacitor parallel to a polyester capacitor plus a parallel resistor with around 100 Ω resistance. That, later, was replaced with the solid ground everywhere. That is the system most Control and PLC systems use today.

The legacy from all these "Thou shalt have No other Ground than One" or "Thou shalt have No Ground" remains and are hailed as The Only Truth in many quarters. In our times of limitless promiscuity, the Ground Everything and Everywhere sometimes knoen as Bonding and grounding is the system that works best in most cases. But it doesn't mean Ground your supply voltages.

You simply have to decide yourself. And be prepared to meet many High Priests that tell you The Truth - different as these truths may be. Still, it is your own decision.

But grounding the -15 V is not recommended by anyone. That is what IRstuff said.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[djs]

My thoughts on this subject are from a different perspective. As a systems integrator, we usually have a large number of analog 4-20ma signals, both receiving and sending. My rule of thumb is to ground the shield at one end only and that is at the receiver end. The reason for this is that the shield then becomes an extension of the ground at the receiver, and any "noise" between the transmitter and the receiver will be shielded by the grounding of the receiver end.

We usually do not ground the analog signal DC supply unless we have issues. Then a single jumper from V- to ground. Also all grounded (non-isolated) transmitters are isolated optically in the control panel.

 

[IRstuff]

I've never dealt with 4-20mA systems, but digital differential receivers can be damaged from power-on transients, IF, and only IF, the sender and receiver are not on the same power system, such as when one side is powered from a grounded AC/DC supply, while the other is powered from a grounded DC inverter, but floating AC/DC supply. This was one case we had to ditch the single-point grounding and strapped the two disparate ground planes together.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[zappedagain]

I'll give another plug for Kimmel & Gerke's book on this subject -

https://www.emiguru.com/books/edn-designers-guide-paypal/

You'll find Chapters 9, 10, and 12 helpful for figuring out cabling.

Z

 

[itsmoked]

123;
Best solution - Get those control wires the he!! out of the same duct as the noise makers.

Second best solution - Do something about the noise generators. Perhaps put some chokes on the power cables to carve off the fast transitions radiating down the cable skins. Use some big 'beads' on them.

Third best and directly next to The Field-of-Woe is connect the shields at the controller and hope djs's argument pays off in your application.

Tree of Woe



Keith Cress
kcress -

http://www.flaminsystems.com