[SI-LIST] : EMC on large distributed systems

Georg Ramsch (georg.ramsch@po.uni-stuttgart.de)
Wed, 14 Jul 1999 11:10:30 +0200

This is a multi-part message in MIME format.
--------------1800EFCC03A10CD3936372C8
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Hello, All !

A friend of mine has a serious problem regarding EMC/SI within a large computing
center.

Spurious data connect problem pop up and dissapear intermittent without the
chance to do some analysis.
The computer systems are conected via hubs/routers/switches, type of connects
are: HIPPI,Ethernet,SCSI,ATM, Fibrechannel.

In addition unused, unplugged older cables remain in the raised floor after the
disassembly of the
computersystems, which belonged to them (were not removed, because if you try
to pull one certain cable out of a bunch then you may damage active cables or
disrupt connectors).
Estimated amount is about some 10th of kilometers representing some 100s of kgs.

The discussion up to now is as follows (in brief): any comments are welcome!

1) What may be the influence of unplugged cables (mostly shielded) running under
raised floors,
which might pick up noise energy and conduct it somewhere else?

2)Regarding switches: through data cables (like HIPPI, SCSI) the cabinets and
GND levels of the attached devices
are electrically coupled. So if you have a GND shift / GND bounce in
one of the devices, this bounce may propagate like a wave through the whole
system of attached
devices and perhaps trigger HW faults (normally switching treshold is about
1.5 V and 0.8 V are the budget for GND shift of the switching transistor and
additional 0.7 V is the noise budget !). True or false ?

3)EMC measures on single devices: influence of devices attached within the
system, where closure/cabinet and GND
of the electronic circuits are connected (some product designer do so to
improve EMC/EMI behaviour)?

All GND levels of all devices would then be connected via one or more
"entrance point" to the overall
chassis structure (assumed, that the cables connect the GNDs of the attached
devices).
Given that case the structure of attached devices would perform a very large
antenna, would it susceptible to short wave radio communication f.e.?
Influence of radio transmitters in the nearer area ??

3) A try (if possible): shorter cables are always better, as they reduce latency
and inductive
path between devices. Often cables are used which are premanufactured to
standardized lengths, but those
length are not needed.

4)Air conditioning effects: if the humidity goes down below 30% rel., then the
airflow in big cabinets
may cause local charge fields, which may discharge thunderstorm-like
(story from an air conditiong engineer responsible for the ac systems at a
computing center).
How long does it take to build up those charge fields, given a swing within
the rel. humidity ?

>>> At the moment we do not regard this as a reason, but pros and cons are
needed.<<<

(A similiar effect is reported from the cleaning of bins of oil tankers; here
the water hoses have a GND
connect at the tip to avoid the buildup of charges on the walls. The discharge
could provoke explosions
of the air/gas mixture inside the bin. Ref.: Jearl Walker / The Flying circus
of physics.

Thanks in advance

CU :-)

Georg
--------------1800EFCC03A10CD3936372C8
Content-Type: text/x-vcard; charset=us-ascii;
name="georg.ramsch.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for Georg Ramsch
Content-Disposition: attachment;
filename="georg.ramsch.vcf"

begin:vcard
n:Ramsch;Georg
tel;cell:0172 8994634
tel;fax:+49 711 6877145 (G3 Fax)
tel;work:+49 711 6875062 (ISDN)
x-mozilla-html:FALSE
org:NEC European Supercomputers
version:2.1
email;internet:georg.ramsch@po.uni-stuttgart.de
note:Harware Support for NEC Supercomputer
adr;quoted-printable:;;c/o RUS=0D=0AAllmandring 30a;Stuttgart;BaWu;70550;D
x-mozilla-cpt:;19904
fn:Georg Ramsch
end:vcard

--------------1800EFCC03A10CD3936372C8--

**** 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 ****