On Sun, 19 Feb 2012 13:33:35 -0600, Tim Wescott
<tim@seemywebsite.please> wrote:
>On Sun, 19 Feb 2012 11:48:19 -0700, Jim Thompson wrote:
>
>> On Sun, 19 Feb 2012 12:15:37 -0600, Tim Wescott
>> <tim@seemywebsite.please> wrote:
>>
>>>On Sun, 19 Feb 2012 09:36:09 -0800, John Larkin wrote:
>>>
>>>> On Sun, 19 Feb 2012 11:19:17 -0500, Jamie
>>>> <jamie_ka1lpa_not_valid_after_ka1lpa_@charter.net> wrote:
>>>>
>>>>>John Larkin wrote:
>>>>>
>>>>>> On Sun, 19 Feb 2012 02:39:53 -0800 (PST), ice <iceeyes82@gmail.com>
>>>>>> wrote:
>>>>>>
>>>>>>
>>>>>>>hi all
>>>>>>>
>>>>>>>system: 9600bps rs485 half-duplex with single master and multiple
>>>>>>>slaves; both master and slave connector have 3 pins (A,B,RTN) cable:
>>>>>>>about 500mt shielded cable with 2 twisted pairs inside, one pair
>>>>>>>used for A/B signals other pair used for RTN
>>>>>>>
>>>>>>>the system works well
>>>>>>>
>>>>>>>questions:
>>>>>>>1. the shield of the cable is connected to earth ground
>>>>>>>(yellow-green wire) at master side only: is this right?
>>>>>>
>>>>>>
>>>>>> It's better to chassis-ground the shield on both ends. That prevents
>>>>>> EMI and ESD and crud from being coupled into the signal conductors
>>>>>> on the end where the shield is floating.
>>>>>
>>>>> John, this is where I must disagree with you. The above assertion is
>>>>>incorrect!
>>>>
>>>> It's perfectly correct. Imagine a long shielded cable with the shield
>>>> open on one end. Induce a nasty current spike, EMI or ESD, into the
>>>> shield at the open end. That will make a big voltage spike in the
>>>> shield... it's far away, many wavelengths and many ohms and many
>>>> microhenries, from earth ground. That nasty voltage spike is directly
>>>> coupled into the signal conductors.
>>>>
>>>> And driving the signal lines will induce voltages into the ungrounded
>>>> shield, radiating EMI. Bad news all around.
>>>>
>>>>
>>>>
>>>>> You DC connect only one end, the other, you can use a small signal
>>>>> cap
>>>>>if it makes you feel better, but never tied two different devices
>>>>>together via their shields on a ground path.
>>>>
>>>> I almost always do. And my stuff works.
>>>>
>>>>
>>>>
>>>>> Argue all you want, I've been there many times over the years, it's
>>>>>basic 101 practice.
>>>>
>>>> In other words, hearsay and superstition.
>>>>
>>>> If you look in a lot of equipment, the grounding of
>>>>>such transmission lines are cap coupled to the chassis, not direct and
>>>>>for good reason.
>>>>
>>>> Disagree, sorry. Shielded Ethernet is grounded on both ends. So are
>>>> standard shielded RS232 and VGA cables and RCA audio/video, and lots
>>>> of other stuff. You *want* ground loop current in the shield, because
>>>> it induces compensating common-mode voltages into the differential
>>>> signal conductors. The cable acts like a balun at higher frequencies
>>>>
>>>> Why leave one end of the shield open, then add caps to local ground?
>>>> It's better to just ground it.
>>>>
>>>> I'm having this argument with a major customer now. I'm selling him
>>>> lots of boxes that interconnect over shielded cables. I'm earth
>>>> grounding all the PC board ground planes with lots of spacers to the
>>>> chassis, and also grounding the cable shields at both ends. They want
>>>> me to float the boards, add an R||C to ground at every board, and
>>>> half-ground the shields. I'm not gonna do it.
>>>
>>>Well, IMHO, you're both wrong.
>>>
>>>Partially.
>>>
>>>And you're both right, partially.
>>>
>>>If you have a setup where the various boxes are fairly close together,
>>>the grounding is thought out well enough to allow for ground-loops in
>>>the shielding, and the EMI environment isn't so severe that you'll melt
>>>the shielding, then having the shielding stay continuous from end to end
>>>is a good thing.
>>>
>>>Presumably this is the world that John is used to.
>>>
>>>If you have a setup where the cable runs are long and ground loops are
>>>uncontrolled, then connecting the ground of one machine to that of
>>>another two hundred feet away may not just be a bad idea from a signal
>>>perspective, it may start a fire.
>>>
>>>Presumably this is the world that Jamie is used to.
>>>
>>>I've seen different grounding schemes used to varying degrees of success
>>>-- each one has its place, many of them can work well, and few of them
>>>work well when you start mixing and matching.
>>
>> I've had to contend with 60VAC gnd-to-gnd high-rise
>> building-to-building.
>
>I've had to contend with 40VAC ground to ground on one factory floor, on
>plugs that were ten feet apart -- and it wasn't a factory with honkin'
>big 500HP motors, either, it was just a moderate-sized light industrial
>building, albeit one big enough (as we found out) to have two separate
>electrical systems.
>
>Fortunately all that we burnt up were some fairly easily replaced test
>fixtures -- but I learned to only trust the "ground all cases together"
>mantra if you had control over what those cases plugged into.
>
>Actually, I had that in mind when I was responding to the post: the AC
>potential (and it was low impedance, too -- it wasn't just a "connect 'em
>together and be happy" thing) between grounds bolsters Jamie's position,
>and the system under test used a grounding system that included all the
>cases being electrically tied together, to bolster John's. Hence -- the
>best answer depends on the circumstances.
Sometimes you need a serious ground strap (think welding cable)
between chassis to force the grounds closer together. The resistance
of the strap has to be much less than the total resistance of the
ground returns in the power system.
When things get extreme, big difference at low impedance, you need
isolation in the signal circuits, or fiber.