RE: [SI-LIST] : FCAL DB9 cable shield

About this list Date view Thread view Subject view Author view

From: Knighten, Jim L (JK100005@exchange.SanDiegoCA.NCR.COM)
Date: Thu Nov 18 1999 - 08:58:24 PST


Chris,

Ah! This is the fun part of EMI.

You correctly characterize my underlying assumption , i.e., the noise source
is imbalance in the differential signal within the cable. Over the past
couple of years, some have termed this "signal-induced" EMI, rather than the
EMI source that is external to the cable and merely "hitches a ride" on the
cable shield. I can point you to at least 2 papers in the August 1999 IEEE
EMC Symposium and one in 1998.

In my experience at NCR, with the FC applications that we have, this has
always been the dominant problem. I must say that we spend a lot of energy
designing our boards to be quiet.

Of course there is one solution that eliminates the ground loop noise
problem (Ron's opinion). That is the use of fiber optics instead of copper
cables. Be warned, however, that this does NOT eliminate the EMI problem.
It merely changes its nature. I have stated in another thread that a myth
of modern technology is that fiber optics are EMI-proof. If you log onto
the emc-pstc list server, you will find that modern high data rate fiber
optics present huge EMI challenges because the electronics used are so noisy
and connectors are so poorly designed for EMI containment.

Jim
________________________________________________________
Dr. Jim Knighten e-mail: Jim.Knighten@SanDiegoCA.NCR.com
<mailto:Jim.Knighten@SanDiego.NCR.com>
Senior Consulting Engineer
NCR
17095 Via del Campo
San Diego, CA 92127 http://www.ncr.com <http://www.ncr.com>
Tel: 858-485-2537
Fax: 858-485-3788

***** Notice the Area Code change from 619 *****

                -----Original Message-----
                From: Chris Cheng [mailto:hycheng@3pardata.com]
                Sent: Wednesday, November 17, 1999 5:48 PM
                To: si-list@silab.eng.sun.com
                Subject: RE: [SI-LIST] : FCAL DB9 cable shield

                here is the fun part.
                i think jim's opinion is based on the assumption that the
                noise current comes from the common mode return current
                due to imperfect differential FCAL signals in which
                case you want the return current to experience as low
                impedance as possible and to minimize the loop between
                the common mode component of the FCAL pair and the
                return current through the shield.
                on the other hand ron's opinion is based on the assumption
                that local noise current NOT generated by the FCAL pair
                will loop thru the shield and thus created EMI problem.
                i guess if the common mode component of the shield current
                is slow enough speed then both opinions can be met. however,
                my suspicion is that the common mode component should
                at least contain spikes due to the inbalance edges
                which is inheritantly highspeed.
                so.... what should we do ??? sigh......
                one more thing, dave instone's mail earlier seems to
                suggest that opening one end with capacitor isolation is
                violating the FC-PH3 spec. besides, classic EMI textbook
calls
                for opening on the floating end. i think i only know how to
                float the receiving end but not the driving end in FCAL i/o
                so in a full duplex DB9 cable, i am out of luck.
                more sigh ...........
                chris

                -----Original Message-----
                From: owner-si-list@silab.eng.sun.com
                 [mailto:owner-si-list@silab.eng.sun.com]On Behalf Of Ron
Miller
                Sent: Wednesday, November 17, 1999 10:08 AM
                To: si-list@silab.eng.sun.com
                Subject: Re: [SI-LIST] : FCAL DB9 cable shield

                Jim & Chris
                The ferrite bead is needed on the shield to reduce EMI. It
will absorb
                circulating currents flowing through the shield returning
through the
                system and rack grounds. Since it is absorptive rather than
inductive
                it actually terminates this energy. It should be designed
to absorb power
                at a frequency associated with the edge rates of the signal.
And it shoud
                be a reasonable short at the data rate. An open at one end
will eliminate
                low frequency (<1 Mhz) power chopper ground currents.
Capacitance can
                ensure connection at higher frequencys.
                Ron Miller
                "Knighten, Jim L" wrote:
                Chris,
                In my experience we have provided a direct short from the
shield of the
                connector to the chassis. This works very well and provides
a good
                continuous shield envelope around the differential pair and
their associated
                transceivers.
                Transfer impedance is a measure of outer current to inner
voltage, so the
                lower the transfer impedance is, the better the shield is.
So, I'd suspect
                that the ferrite bead/ac short approach is not as good.
                Jim
                ________________________________________________________
                Dr. Jim Knighten e-mail:
Jim.Knighten@SanDiegoCA.NCR.com
                <mailto:Jim.Knighten@SanDiego.NCR.com>
                Senior Consulting Engineer
                NCR
                17095 Via del Campo
                San Diego, CA 92127 http://www.ncr.com
<http://www.ncr.com>
                Tel: 858-485-2537
                Fax: 858-485-3788
                ***** Notice the Area Code change from 619 *****
                                -----Original Message-----
                                From: Chris Cheng
[mailto:hycheng@3pardata.com]
                                Sent: Tuesday, November 16, 1999 6:02 PM
                                To: si-list@silab.eng.sun.com
                                Subject: RE: [SI-LIST] : FCAL DB9
cable shield
                                jim,
                                  excellent point. one follow up question,
should the
                shield/connector
                                housing be shorted to chasis/local ground
through
                                a) direct short
                                b) ac short through ferrite beads that have
more transfer
                impedance
                                then both the shield and the connector ?
                                chris
> -----Original Message-----
> From: owner-si-list@silab.eng.sun.com
> [mailto:owner-si-list@silab.eng.sun.com]On
Behalf Of
                Knighten, Jim L
> Sent: Tuesday, November 16, 1999 5:35 PM
> To: si-list@silab.eng.sun.com
> Subject: RE: [SI-LIST] : FCAL DB9 cable
shield
>
>
> Fibre Channel cables are used for
relatively short runs (a
                few tens of
> meters maximum). For short cable runs,
ground potential
> differences within
> facilities, hence ground loop currents,
are not an issue
                of concern.
>
> What is an issue of concern is EMI from
these cables. In
                all of our
> applications, EMI is generated by the
signal within the
                cable.
> It is caused
> by differential signal imbalance, which
results in a net
> common-mode current
> on the signal pair. This common-mode
current returns to
                its
> source via the
> inner surface of the cable shield. It
leaks out of the
                cable assembly
> through the inherent leakage of the cable
shield (i.e.,
                the
> surface transfer
> impedance of the cable shield) and the
though the
                connector. In general,
> the connector is the biggest emitter of
EMI. In the case
                of a properly
> shielded connector (conducting backshell
and a 360 degree
                bond to
> the cable
> shield) then the leakage mechanism is the
connector
                shield's transfer
> impedance. This is usually bigger than the
cable shield's
                transfer
> impedance, but is manageable. In the case
of a cable
                shield that is not
> connected to the connector shield, then
the leakage of
                common-mode EMI
> radiation is very high at the connector.
That's why these
                Fibre Channel
> cable assemblies have continuous shields
from cable to
                connector and then
> onto to the chassis wall that contains the
mating
                connector.
>
> As far as Ethernet is concerned, the
specifications
                prohibit grounding the
> shield of the cable to the chassis through
the connector
                (except
> at a single
> location). This is a prime reason why
Ethernet is such a
                horrible EMI
> offender. It presents a highway for noise
internal to a
                chassis
> to ride out
> on the ungrounded cable shield.
>
> Jim
>
________________________________________________________
> Dr. Jim Knighten e-mail:
                Jim.Knighten@SanDiegoCA.NCR.com
> <mailto:Jim.Knighten@SanDiego.NCR.com>
> Senior Consulting Engineer
> NCR
> 17095 Via del Campo
> San Diego, CA 92127
http://www.ncr.com
                <http://www.ncr.com>
> Tel: 858-485-2537
> Fax: 858-485-3788
>
> ***** Notice the Area Code change from 619
*****
>
>
> -----Original Message-----
> From: Christian S. Rode
                [mailto:csrode@mediaone.net]
> Sent: Tuesday, November
16, 1999 4:57 PM
> To:
si-list@silab.eng.sun.com
> Subject: Re:
[SI-LIST] : FCAL DB9
                cable shield
>
> Certainly any difference in
ground potential
                between
> locations could
> cause
> current to flow
preferentially through the
                shield. I don't
> think being
> on
> different phases is as much
the issue as
                unbalanced use. If
> one phase
> of
> your home triplex service is
used more
                heavily or a large
> motor blows a
> phase on three-phase
industrial service I
                think the
> center-ground taps
> will create a voltage
differential.
>
> Capacitively or inductively
coupling shields
                and ground at
> one end
> solves the ground current
problem but you'll
                have to choose
> a
> coding scheme without a DC
component...
>
> Isn't this why optocouplers
were invented?
> Aren't there safety issues,
too? Seems I
                remember that it's
> not to
> (current?) code to run
Ethernet cabling
                between buildings.
> Not that
> people don't do it anyway...
>
> > >
> > Wouldn't this cause nasty
galvanic
                currents through the
> shield if the two
> > pieces of equipment,
attached by the cable
                with the DB9s,
> were, say, on two
> > different phases of a
three phase power
                feed to a
> building.
> >
> > Am I missing something?
> >
> > thanks
> >
> > Tom Gandy
> > Industrial Catalyst
> >
> > **** To unsubscribe from
si-list: send
                e-mail to
> majordomo@silab.eng.sun.com. In the BODY
of message put:
                UNSUBSCRIBE
> si-list, for more help, put HELP. si-list
archives are
                accessible at
> http://www.qsl.net/wb6tpu/si-list ****
>
> **** To unsubscribe from
si-list: send
                e-mail to
> majordomo@silab.eng.sun.com. In the BODY
of message put:
                UNSUBSCRIBE
> si-list, for more help, put HELP. si-list
archives are
                accessible at
> http://www.qsl.net/wb6tpu/si-list ****
>
> **** To unsubscribe from si-list: send
e-mail to
> majordomo@silab.eng.sun.com. In the BODY
of message put:
> UNSUBSCRIBE si-list, for more help, put
HELP. si-list
                archives
> are accessible at
http://www.qsl.net/wb6tpu/si-list ****
>
>
                                **** To unsubscribe from si-list: send
e-mail to
                majordomo@silab.eng.sun.com. In the BODY of message put:
UNSUBSCRIBE
                si-list, for more help, put HELP. si-list archives are
accessible at
                http://www.qsl.net/wb6tpu/si-list ****
                **** To unsubscribe from si-list: send e-mail to
                majordomo@silab.eng.sun.com. In the BODY of message put:
UNSUBSCRIBE
                si-list, for more help, put HELP. si-list archives are
accessible at
                http://www.qsl.net/wb6tpu/si-list ****
                --
                Ronald B. Miller _\\|//_ Signal Integrity Engineer
                (408)487-8017 (' 0-0 ') fax(408)487-8017
                     ==========0000-(_)0000===========
                Brocade Communications Systems, 1901 Guadalupe Parkway, San
Jose, CA 95131
                rmiller@brocade.com, rbmiller@sjm.infi.net

                **** To unsubscribe from si-list: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list, for more help, put HELP. si-list archives are accessible at
http://www.qsl.net/wb6tpu/si-list ****

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:38:59 PST