[SI-LIST] Re: backplane PDN/EMC challenges
- From: steve weir <weirsi@xxxxxxxxxx>
- To: qantrix <qantrix@xxxxxxxxx>
- Date: Wed, 23 Feb 2011 22:40:59 -0800
Qantrix, please review my first reply.
Steve.
qantrix wrote:
> Hi Steve,
>
> thank you. I got the point about dc return topology.
>
> Now about EMC.....
>
> I consulted with people doing big 10G+ routers in company that can
> afford long investigations and different layouts experiments ( most
> of designers working in small companies, like i as well, can
> experiment from design to design only , unfortunately...). So the
> answer was - connect all them together! No splits, no gnd cuts,
> moating etc. only high performance a/d or pll need some isolation or
> single point gnd.
>
> I have read free list archive about this question - and there is a lot
> here , also i have Henry Ott, Paul's , Archambeault and other books
> , some books can be partially read on google books.Reading all this
> stuff i understood that there is no one solution - there is single
> point ground people, moating ( quite io ground concept) people and
> gnd plane bonded to chassis(as many points as better, around all the
> board - the best case) people.
>
> All books have similar pcb example - low speed io connector are
> moated, high speed are bridged from gnd plane point. It looks like
> some EMC dogma that can not be omitted by author...
>
> But what about "Real world EMI control" "by Design" ?
>
> I spoke now about PC grade equipment( in my case it digital video
> router that will take video from PC or other source)
> pcb with non isolated (from dc point) dgnd is required, no medical
> requirements etc..
>
> Yesterday i opened Apple G5 PC ( the point was that regular pc is "low
> end"device that sold in millions and surviving on the edge from emc
> point because of cost reduction. G5 - looks like more professional and
> it built like that ). So first Fluke test result was - chassis is
> direct connected to DGND. Now what is going on in internal layers -
> thats the good question!! Maybe there is moating or something like
> that. Also we didn't see any groups of 0R resistors on the pcb edges
> (usually connecting between gnds in some designs). Graphics card vcc
> is power via common mode choke!
> Now we have one design (the box is small - like 15x15 cm) - there is
> no connection between CHGND and DGND, cut is done under differential
> lines! going out of device to cable. Now we have 0r resistors to
> connect between them . EMC test shows that resistors are connected
> making worst, and device passed class B with good marging. I say again
> - there is gnd plane cut under highspeed differential pairs!!! Some
> thing that prohibited even to think about.. :)
>
> From other point we have designs ( 19 inch rackmounts) where
> unconnected gnd cause for EMI disaster. Only connection back 0R
> resistors make device passing EMC test.
>
> As i said before i asked people from GBs ethernet world and from
> military video world guys - they are doing gnd plane without any
> separation bonded to chassis as good as possible.
>
> Where is the truth?????
>
> I feel like the bonded dgnd to chassis - is best and simplest way....
>
>
> Thanks in advance!
>
>
> 2011/2/23 steve weir <weirsi@xxxxxxxxxx <mailto:weirsi@xxxxxxxxxx>>
>
> qantrix wrote:
>
> I don't wanted insert common mode choke to highspeed return,
> only for power.
> i uploaded simple illustration here
> http://img526.imageshack.us/i/17449681.jpg/
>
> That topology is exactly what I was cautioning you about.
>
>
> those planes will have coupling to highspeed return and
> produce common mode noise on + and - and cmc will cut it.
>
> The chokes will impose impedance between the power supply inputs
> and the power supply. If for example, your power cables run
> external to the box and are the major antennae in the system, then
> CM chokes can do you good. If however, you have the more common
> situation that the power source has a common connection to the
> case, even if DC isolated with a capacitor, then you've set-up a
> resonant system between your signal common and the enclosure.
>
> Now if we don't use cmc , just route power return as a star.
> Common point of connection of power return and signal gnd will
> be at power entry point.
>
> Star routing is good for low frequency analog. It is usually a
> disaster for high speed digital and EMC>
>
>
>
> About safety - you mean isolation between chassis to sgnd
> from ground loops currents or high voltage isolation? We don't
> need high voltage isolaition.
>
> No, I mean current crowding in conductors. There are situations
> where one has to prevent signal common from becoming a significant
> current path for one side of the power supply.
>
>
> 2011/2/23 steve weir <weirsi@xxxxxxxxxx
> <mailto:weirsi@xxxxxxxxxx> <mailto:weirsi@xxxxxxxxxx
> <mailto:weirsi@xxxxxxxxxx>>>
>
>
> Inserting magnetics between Vss used as high speed signal
> return
> and power common that returns to chassis creates lots of
> opportunity for resonances in the 100MHz and higher range
> due to
> the shunt capacitance between the Vss planes, the power
> common and
> the chassis. A CM choke can aggravate DC current crowding and
> sneak path through signal common connections. DC isolation
> between signal Vss and power supply common requires isolated
> voltage conversion.
>
> The bottom line is you need to determine what is required
> to meet
> applicable safety standards for your equipment. If those
> can be
> satisfied with common signal and power commons, then the
> simpler
> arrangement of DC connection between power and signal return
> almost always leads to fewer and more manageable problems.
>
> Steve.
>
>
> qantrix wrote:
>
> Hi All ,
> we are challenging how to provide power/return to cards
> connected to
> backplane
> and how to deal with chassis/signal ground
>
> So device will transfer high speed signals with Tr=0.1ns.
> Digital video data
> rates
> General curretn consumption will be 12V 40A. So like
> 5amps for
> each of
> cards(8).
> Device doesnt have high voltages or high energy voltage
> sources. general
> digital video application.
>
> There is two options - do separate polygons for each of
> VCC_1...8 and its
> returns by using 10A common mode choke placed at power
> source.
> Doing that supply and return currents will be under control
> and no return
> current hot spots will be produced. and do it for all
> plug in
> cards.
>
> Second option is doing common 12V polygone and return
> share with
> signal(highspeed ) gnd - there will be a lot of gnd
> planes in
> 14 layer
> stackup.
> On plug in cards also - can be two strategies - separate
> polygones(control
> for supply/return currents) or common - current will
> return
> with signal
> return in connector)
>
> First option is rigth from general understanding point.
> But in
> practice - do
> the second variant work well? What says experience?
>
> Another question - what is common practice for chassis
> gnd in
> rackmount
> devices - does in need to be dc separated by caps from
> sgnd?
> Many people suggest to not separate at all - that will
> give
> good emc
> perfomance - all noisy gnds will be bonded with minimum
> inductance to metal
> box(faraday cage) also esd current will have low impedance
> return path.
> Other people say that noisy gnd planes will drive
> connectors
> area and cables
> will act as efficient antenna.
>
> I'm betweeen two opinions, by my self i think common
> plane for
> two problems
> is good and simple way to go.
>
>
>
> ------------------------------------------------------------------
> To unsubscribe from si-list:
> si-list-request@xxxxxxxxxxxxx
> <mailto:si-list-request@xxxxxxxxxxxxx>
> <mailto:si-list-request@xxxxxxxxxxxxx
> <mailto:si-list-request@xxxxxxxxxxxxx>> with 'unsubscribe' in
>
> the Subject field
>
> or to administer your membership from a web page, go to:
> //www.freelists.org/webpage/si-list
>
> For help:
> si-list-request@xxxxxxxxxxxxx
> <mailto:si-list-request@xxxxxxxxxxxxx>
> <mailto:si-list-request@xxxxxxxxxxxxx
> <mailto:si-list-request@xxxxxxxxxxxxx>> with 'help' in the
>
> Subject field
>
>
> List technical documents are available at:
> http://www.si-list.net
>
> List archives are viewable at:
> //www.freelists.org/archives/si-list
> Old (prior to June 6, 2001) list archives are viewable at:
> http://www.qsl.net/wb6tpu
>
>
>
>
> -- Steve Weir
> IPBLOX, LLC 150 N. Center St. #211
> Reno, NV 89501 www.ipblox.com <http://www.ipblox.com>
> <http://www.ipblox.com>
>
>
> (775) 299-4236 Business
> (866) 675-4630 Toll-free
> (707) 780-1951 Fax
>
>
>
>
>
> --
> Steve Weir
> IPBLOX, LLC 150 N. Center St. #211
> Reno, NV 89501 www.ipblox.com <http://www.ipblox.com>
>
> (775) 299-4236 Business
> (866) 675-4630 Toll-free
> (707) 780-1951 Fax
>
>
>
--
Steve Weir
IPBLOX, LLC
150 N. Center St. #211
Reno, NV 89501
www.ipblox.com
(775) 299-4236 Business
(866) 675-4630 Toll-free
(707) 780-1951 Fax
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List technical documents are available at:
http://www.si-list.net
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
