Machsupport Forum
Mach Discussion => General Mach Discussion => Topic started by: stirling on February 07, 2014, 05:07:48 AM
-
In every off the shelf parallel and serial cable I've looked at, both end shells are connected via the shield. I've just double checked my PC and printer and was not too surprised to find that the port receptacles on BOTH were connected to their respective chassis which of course in turn are both grounded. So it seems that off the shelf kit does not follow this ground shield at one end only philosophy.
And yet - when I look it up on tinternet - there it is all over the place - ground shield on signal wires at one end only....
Anyone care to shed some light/wisdom/knowledge/thoughts on this?
EDIT: Just to be clear, I understand the logic behind the one-end-only. My question is why doesn't this appear to be the case in "off the shelf" kit.
-
Shielding is done / required / utilized and even custom designed for a lot of different reasons, but, simplisticaly:
- to keep what is inside, inside the shield
- to keep what is outside from interfering what is inside
So now the question becomes...... What are you trying to control with the shield?
Using a crude water analogy... water may be on a pipe. Where do you want the water to go ?
Go back to the sanitation sewer, air dry, or go to a designated sump which makes it easier to get into the sewer line or a designated sump system, just contain it ( ie; tank to tank ) or even a combination of two or three of them.
So grounding of the shield could be subjective and only the person looking at the big picture for what was considered may know why
what is being done with the shield is appropriate. But generalizations do come about and are not necessarily always true.
From a non-electrical ....and may be all "wet", just some thoughts,
RICH
-
Just my 2p’s worth…
Basically, in most countries, regulations for mains connected equipment provide that unless a product meets the ‘double insulated’ (or a couple of other) standards all metal parts that we may touch must be bonded to protective Earth.
With our CNC the LPT pins 18-25 are used as ‘signal GND’ and sooner or later, in our build, we connect them to Earth - this is where the potential problems with electrical noise start if we don’t follow the one ended connection rule for shields / screens.
Tweakie.
-
The resident electrical expert at CNC Zone, Al The Man, has been posting information for the last few years that says the currently accepted method is to ground shields at both ends.
-
I can tell you that the Allen Bradley drives I use say to connect the encoders shield at both ends and I have done that on some and not on others and never noticed a difference but I am talking normal incremental 5v square wave differential encoders.
I also have some motors with sick stegman sine/cosine encoders and they will just not work without a shield or the shield only connected one end, they need the shield connected at both ends.
Now I have seen all sorts of explanations of how to shield and where to connect and some of the explanations have mentioned that it is frequency dependant. I have always tended to connect one end only as that seemed to be current thinking but like everything else, thinking changes and you can never get a clear answer.
I say connect one end and if you have problems try connecting both, if you still have issues then start crying (would pull my hair out but its already gone :D )
Hood
-
Current Thinking?
The SmoothStepper was a good example of this, some people suffered from noise problems and some like me never. Well thats a slight lie as I did slightly on one machine when I whacked the E-Stop it would occasionally screw up the SS. So someone found their probs were halted when they remove a resistor and cap that connected the SS Gnd to the shield of the cable and to the computer, I tried that and it cured my very occasional issue. Greg started sending out SS's without these components and some people still had issues with noise, so resistor and cap were soldered on and their problems stopped.
So why is it it can cure for some and make worse for others? I say there is no hard and fast way to stop the noise getting in and it is all dependant on the rest of your system, so one end may be good on one system where another needs both but I can say for sure no shielding will be the worst.
Hood
-
From what I've gathered, it's all about the "loop".
If every component of a system is grounded properly, connecting the shield at both ends would drain well.
But due to improper grounding, the drain can act as a path to ground for other components, not just the emi.
One end only is probably best in the "hobby" world.
Professional builds in the industrial world may be different.
AlphaWire is king in this dept.
http://www.newark.com/pdfs/techarticles/alphaWire/USC.pdf
Russ
-
... to reinforce what I've gathered ....
This problem is very prevalent in audio systems.
"One end" connected actually acts as an antenna.
Interesting writing:
http://www.equitech.com/support/wiring.html
(from above link) In virtually every case, with balanced power, balanced audio can be interconnected in a standard way with the shield hooked up at both ends. The only exception is when a piece of gear has a "dirty" chassis which requires isolation away from the rest of the grounding system. This can usually be accomplished by lifting the audio ground at the input and isolating the offending chassis from other chassis with insulators. A "dirty" chassis condition is rare in professional audio equipment and it often is the result of a substandard power supply or the audio ground not being connected to the chassis. These problems can often be fixed with some effort. In general, wiring and grounding techniques are far simpler with balanced power and it is easier to identify and deal with any offending piece of equipment.
-
I think sometimes it pretty much boils down to some twisted form of black magic. I think it has a lot to do with the location of the component generating the noise. Envision a control cabinet with nothing but low voltage/current components like a PC and BOBs. Leaving that you have a few control wires going to a VFD. I would likely start by breaking convention and grounding the shield at the VFD end only. Most I have seen in industrial controls suggest grounding the control end only but I view that as a suggested starting point. If that doesn't work, try grounding both ends. If that doesn't work try grounding only the end opposite of the controller. Sometimes floating the ground of a power supply helps, others it causes a real mess. I think it boils down to the fact that we simply do not know enough about EMF. Some of it can be extremely difficult to measure and it is hard to analyze data you do not have. EMF is like a fire. Try to contain it as close to the source as possible because it is going to do everything it can to spread to everything in sight.
Most of the time you can narrow it down to the source of the problem pretty quick. Other times it takes for ever and others you never do. The ones that just blow my mind is when you have a machine all wired up and tested and maybe even run for extended periods of time with no issues. Pick it up and move and now noise is a serious problem. Everything I know to measure are dang near identical. Dedicated ground rods megged to within a few ohms of one another. Same wiring, same routing, same input power, same type building, same kind of lighting, relative humidity........ every thing I can think of. Obviously I'm missing something but what? The only thing I can come up with is that it is simply located at a different spot on this molten Iron core rock we call earth. Could drive a man mad. Most of the time the solution is to fiddle with the shields again. I just find things like this bizarre to say the least and anywhere from humbling right up to embarrassing. ;D If any of you guys have seen similar and actually narrowed it down and could put your finger on it how about share it. Or share the most bizarre source of noise you have found and the solution.
Brett
-
most bizarre source of noise you have found and the solution
Blinking Christmas decorations.
Just need to waite until after the new year. ;)
RICH
-
Some thoughts that come to mind.....
some of the explanations have mentioned that it is frequency dependant.
Yes,true if you want the shield to act as low inpedence to some group of frequencies, ie;
it could act as antenna which is tuned to a group or actualy be a high impedance to a frequency and
act a short....but...depends and that's were the famous word "subjective" comes into play
resistor and cap
a resistor and cap ie; rc circuit ( charge up and discharge at a time rate ) can be tuned and act as wide or narrow filter to ground
black magic
Yes it certainly seems that way since I have known some sharp cookies who realy couldn't explain why something worked
or didn't work
Location / proximity to the source, intensity of the signal, frequency, shield quality, ground quality, type of ground system
etc all have an affect on what may or may not happen. One of the problems with all this is that one can't see it or if it is seen it's via a device which shows what your interested in in some manner.
RICH
-
But after all this I'm back at my original question which was why commercial ready made up cables have the shield connected at both ends. I guess it's either that or just one end so the designer tossed a coin and said - two it is ;)
By having both ends connected from the factory, it gives you the option as there is an "interrupter" unit that can be installed between two cables to "break" the loop if necessary.
fwiw
Russ
:)
-
But after all this I'm back at my original question which was why commercial ready made up cables have the shield connected at both ends. I guess it's either that or just one end so the designer tossed a coin and said - two it is ;)
Probably because the ports are tied to the PC power supply ground, which ties into the AC ground. So you typically have good grounding at both ends. (In a typical PC application)
-
F***ing hell - sorry Russ and Gerry - I just managed to delete my last post (which you've just quoted). How do I put it back?
EDIT: Bollocks to it - here it is: ;D
Thanks guys - interesting stuff.
FWIW - This is what I've picked out of it all
1) If your grounding is spot on then ground BOTH ends of your shield.
2) If your grounding is not spot on then grounding both ends is likely to introduce earth loops so just do one end.
3) but in the real world it's not that easy so if it doesn't work then do something else.
But after all this I'm back at my original question which was why commercial ready made up cables have the shield connected at both ends. I guess it's either that or just one end so the designer tossed a coin and said - two it is
-
I reckon you got it Russ - thanks. Gerry - good point but it's not just PC cables - I have a Hypertherm plasma interface cable right here that is screen-grounded at both ends.
-
actually called "Isolators"
Although ground loops often involve power line safety ground connections, disabling them is both highly dangerous and illegal.
However, devices called “ground isolators” can be inserted in the signal path to break the loop safely.
Interesting ... : a transformer , neat.
http://www.prosoundweb.com/article/pesky_ground_loop_problems_a_look_at_solving_them_the_right_way/
-
http://www.cnczone.com/forums/general_electronics_discussion/177819-computer-numeric-control_grounding_bonding_article.html
-
My post's are audio related, but I think the principals are the same (or close).
One thing about the audio, it's easy to see (hear) the emi at almost any level. Otfen far before it would become a problem with a 5v digital pulse.
Studying audio makes it easier for me to visualize ...... (huh, visualize audio ? .. I think you know what I am meaning) I need all the help I can get.
Russ
-
Thanks again guys - meanwhile another (related) question: When we talk about noise - we're talking about induced voltages right? Any ideas on what levels we could be talking about? i.e. what sort of voltages can be induced into a screen?
-
Just a clarifcation of what noise is:
Well use Russ's audio analogy since you can hear it.
Noise is just many frequencies over a frequency range which we don't understand. You can hear a single tone or some combination
like an orchastra playing and make sense of them. But all that hiss or crackling is just noise. They both exist and depending on the amplitude of one to the other one may dominate.
The same applies to higher and lower frequencies out of our hearing range.
A good example of noise would be sparking from brushes of a motor which puts out high noise levels ie; many random frequencies
through out the spectrum . Should see what happens when tig welding welding is done and a computer is on and near by.....
heck wiped my friends hard drive.
So if you cut a energy field, say one with a whole bunch of noise, with a wire, a flow of noise electrons will occur on the wire and you can measure the voltage. The noise was inductively coupled from a radiating source. So if the noise level is high compared to the
a signal that you were sending on the wire, the noise would dominate and say a device recieving it could not process it.
Sorry for the crap explainations as most of the tech stuff is saved to memory and just can't seen to find the folder it was place in! ;)
What do you mean by ...induced into a "SCREEN".
RICH
-
Thanks again guys - meanwhile another (related) question: When we talk about noise - we're talking about induced voltages right? Any ideas on what levels we could be talking about? i.e. what sort of voltages can be induced into a screen?
If you know what the source of the noise is you may able to get a good idea here.
http://edn.com/electronics-blogs/bakers-best/4371301/EMI-problems-Part-three-strength-of-EMI-radiated-signals
Brett
-
Thanks again guys - meanwhile another (related) question: When we talk about noise - we're talking about induced voltages right? Any ideas on what levels we could be talking about? i.e. what sort of voltages can be induced into a screen?
It's not hardly that simple.... There are many types of noise, and many ways for it to cause problems. One of the purposes of a shielded cable is to provide a constant impedance over the length of the cable. A break in the shield causes a change in impedance, which can cause the signal to be reflected back to the source, even in the absence of any external noise sources. Noise can couple capacitively, magnetically, or electrically. It's a very complex business. Working with 5V signals, it's not terribly diffcult to get induced noise of the same magnitude, or greater, than the signal itself.
That said, you really should not havve any problems it you follow some pretty basic good grounding, and routing, practices, and keep the signal impedances as low as possible, by using stiff pull-up resistors. A good shield, connected at the source end only, should work just fine, except in unusual cases. Connecting them at both ends *can* work better, but only if the entire system is well-understood. In the typical hobby CNC case, it's not, beause you don't know how things are connected inside the power supplies, motor drivers, BOBs, etc. Connecting at one end will avoid a lot of potential ground loops.
My machines are wired with shields connected at the source end only. All my limit and home switches, and other signals are 5V only, but with 300 ohm pull-ups. I even run the servo motor cables and their encoder cables side-by-side, actually Zip-tied to each other about every 12 inches, for almost 20 feet, and I've never had a single noise-related problem.
Regards,
Ray L.
-
But after all this I'm back at my original question which was why commercial ready made up cables have the shield connected at both ends. I guess it's either that or just one end so the designer tossed a coin and said - two it is
Ian,
I just checked a couple of USB cables that I have here and the PC Earth is not carried through like it is with LPT cables - with my USB cables it is the signal GND which is carried through.
Tweakie.
-
RICH - understood - but my question was about specifically trying to get a handle on the AMPLITUDE (Vp2p/Vrms/dB) of the noise we might typically see. After all if we're happy to talk about signal level and we're happy to talk about signal to noise ratio then we MUST be happy to QUANTIFY noise in the amplitude domain - yes? I know it's kinda like asking how long is a piece of string - but just trying to get a picture. We hear all the time about "noise" in "general terms" but no-one ever quantifies it. That's my interest in asking is all.
Ray - thanks - but see my answer to RICH.
Tweakie - Interesting you mention USB because I tried to find if the USB spec specified how the SCREEN is connected on those also. All I could find was one link that said the screen is NOT connected at the USB device end. So it looks (though not conclusively) like USB is in the one-end-only camp.
Brett - perfect - exactly what I was after - thankyou.
-
Don't have any numerical values. If i get a chance will blow the dust off the specturm analyzer and show / create an actual example.
But in the mean time, here is an example that was a rather interesting experiment we did a loooooong time ago.
On transmision of low power say 1 watts output from a a transciever so about 6 watts out of the antenna, we would lock up an on line pc. So we put an isolation transformer in the ac line to it, completely enclosed the pc , wring etc, in a solid metal shield, EXCEPT for the face of the monitors glass tube, and provided a separate grounding system for the pc and downstream side of the transformer. The ground wire was actualy a 1" diameter piece of copper tubing and the inside of it was used as the conductor tied to a 10' driven ground rod.
If the transmitted freq was near (don't remember how much) the pc cp operating htz, fundamental overload occured
and the pc screen would first turn blank white and stay that way even after the transmitioin was stopped and pc unplugged
for some time. Placed a wire meshed screen over the monitor screen some distance away from it and tied it to the ground system.
Don't remember the extimated / calculated freq based on screen opening, and it did provide for interference rejection but was not satisfactory, ie; could move the interfering freq a little closer ......signal to noise ratio was down a little.
Here is another un quantified one.
Just do TIG welding near a PC . Say the pc is 15 away, shoot, it wiped the hard drive!
Another little tid bit for info:
Ever read that tag that's on an electronic device that says something like " this device complies with FCC rule ???15"
so the device cannot radiate signals but must be able to accept them.
Another practical example:
House telephones, cheap ones had no filtering for noise or filtering that was about usless, but you would never have a problem with
a good phone made by AT&T. Not much of a problem anymore as the phones function in the giga htz range.
Then there are rules form a safety point of view that came out some 10? years ago. The field as measured in gauss? had to be within a certain level from a people location point of view......don't remember specifics.
So the EMI / noise,contol & elimination of it is not simply a some value that one can use as a quideline.
So one must measure the values to have a feel of relative intensity and that info is tied to how sensitve the offended circuit / electrornics is. Also as distance increases from the source the field intensity drops off.
That's why you don't see any qauntification, all is specific to the situation at hand, and also most folks don't have the equipment to measure.
RICH