Re: [SI-LIST] : High Speed Backplane Connector Recommendations

About this list Date view Thread view Subject view Author view

From: Ron Miller (rmiller@Brocade.COM)
Date: Fri Jan 14 2000 - 10:34:57 PST


Steve

I disagreee with your conclusion. See below.

sweir wrote:

> Bruce,
>
> No, both the EMI and the SI will be worse. To understand this, think about
> how each of the LVDS drivers develops its signal:
>
> The LVDS "high" line sources from a current source in the driver, while the
> "low" line sinks to the driver ground. The driver current source has a
> limited common mode range, as does the sink. At the far end we place a 100
> ohm resistor between the lines. So, ok current flow in one direction or
> another creates a relative voltage difference across the resistor. BUT,
> the common mode voltage is just the average voltage of the two. Our
> receiver must be able to:
>
> 1) Survive the common mode voltage, and

Yes, but anyone with an ounce of sense will still use the same supply voltages
so there should be no problem

>
> 2) Register the intended results
>
> For a demonstration of 1) you can destroy a perfectly good PC and printer
> by lifting the grounds on both and plugging each into extension cords at
> opposite ends of your house, and then connecting the PC to the
> printer. When your refrigerator turns on and off watch as one or the other
> lets all the smoke out.

Again I say this is a fools scenario.

Incidentally, all equipment uses the green wire for chassis/safety ground and the
green is connected at the power box, circuit breaker panel to the low side of the power.
Because of inductances it is possible to get an AC difference up to a couple volts due to
unequal loads on the AC phases at a wall socket. DC is not effected.

Within a system as in a PCB as was the question set up, the ground voltages are very
close to equal at the receiver and driver.

>
>
> This is one of the reasons that box to box transports use transformers or
> optics. If you operate within a box, it is generally much more economical
> to deal with a common ground.

:Yes these work well to eliminate any offset where long cables are used.

>

Conclusion: within a chassis all grounds should be connected together keeping
any common mode voltages low. Then coplanar waveguide works nicely without
a ground plane.

Ron Miller

>
>
> Regards,
>
> Steve.
>
> At 05:43 PM 1/13/00 -0700, you wrote:
> >Steve,
> >
> >What if you put no ground plane, just two traces on a PCB by themselves?
> >Can we not look at these by themselves as a transmission line? If driven
> >differentially and with a well balanced source impedance will not the
> >current flowing down one trace then return down the other assuming they are
> >terminated in their characteristic impedance?
> >
> >You may ask what advantage this would have. It might allow one to create
> >higher impedance PCB transmission lines.
> >
> >It might have the disadvantage of worse EM radiation. But SI should be
> >good, shouldn't it?
> >
> >Bruce
> >
> >----- Original Message -----
> >From: sweir <weirsp@a.crl.com>
> >To: Bruce W. Marler <BMarler@WooshCom.com>
> >Sent: Thursday, January 13, 2000 4:54 PM
> >Subject: Re: [SI-LIST] : High Speed Backplane Connector Recommendations
> >
> >
> > > Bruce,
> > >
> > > When we place two conductors in a diff pair over a ground plane, 85-90% of
> > > the energy in each lead couples to the ground plane, while the remaining
> > > 10-15% couples to the other trace in the pair. This is true whether we
> > > edge couple, or broadside couple.
> > > What this means is that if your ground plane does not provide a good
> >return
> > > path, then you will be subject both to SI and EMC problems you don't
> > > want. The bottom line is that even though you are using diff signals, as
> > > long as they are DC coupled, ( LVDS, BLVDS ), then you need to assign a
> > > good deal of your connector real-estate to signal ground, so that you can
> > > maintain a high frequency return path both for EMC and SI. If you miss
> > > this point, your timing will degrade, and you may end up with unmanageable
> >EMI.
> > >
> > > The AMP stuff is "Z-pack HS3", pricey, but very good. The data sheets are
> > > on their www site. If you can eat the power, you can drive those guys
> >with
> > > Vitesse 870/880's which i/f 2Gbps to FPGA's nicely with only one pair in
> > > the bp. If you are hell-bent on LVDS, there are several solutions out
> > > there to reliably move data at higher rates. Lucent has a four pair
> >ganged
> > > TX/RX module available in their ASIC's which moves 2.4Gbps as a single
> > > stream. You could also consider the channel link parts that clock line
> > > data reliably between 350 and 750 Mbps / pair, 1.4 - 3+Gbps / 5 pair.
> > >
> > > Regards,
> > >
> > >
> > > Steve.
> > > At 04:19 PM 1/13/00 -0700, you wrote:
> > > >Steve,
> > > >What do you mean when you say that "most of the energy still couples
> >between
> > > >the returns and the individual signals in each pair"?
> > > >
> > > >Also what is the more expensive AMP connector that you refered to which
> >can
> > > >transport 2.4 Bbps signals?
> > > >Thanks
> > > >Bruce
> > > >----- Original Message -----
> > > >From: sweir <weirsp@a.crl.com>
> > > >To: Bruce W. Marler <BMarler@WooshCom.com>
> > > >Sent: Thursday, January 13, 2000 12:07 PM
> > > >Subject: Re: [SI-LIST] : High Speed Backplane Connector Recommendations
> > > >
> > > >
> > > > > Bruce,
> > > > >
> > > > > I think your plan has some scalability problems. Multidrop backplanes
> >are
> > > > > generally not a very good thing to use these days as the parasitics
> >place
> > > > > an upper limit which is very limiting, even using a fairly good
> >technology
> > > > > such as LVDS. It sounds like you are planning to use BLVDS on a
> > > > > non-redundant backplane.
> > > > >
> > > > > For connectors, the older and less expensive AMP 2mm hard metric,
> >which
> > > >has
> > > > > several alternate sources is not bad, provided you do your
> > > > > homework. Please remember that most of the energy still couples
> >between
> > > > > the returns and the individual signals in each pair, sic, make sure
> >you
> > > > > have adequate ground density. AMP has a much more expensive connector
> > > >line
> > > > > intended to reliably transport 2.4Gbps across backplanes than the HM
> > > > > stuff. The HM connectors are a better choice if your plan is to
> >remain at
> > > > > a fairly low performance of 100MHz on each wire pair. A single
> >connector
> > > > > provides 125 signal pins in 50MM. Realistically, you can get 48 good
> > > >pairs
> > > > > out of one such connector.
> > > > >
> > > > > I strongly recommend that you get a good SI consultant to review your
> > > > > backplane plan before you commit to your architecture. My concern
> >arises
> > > > > from the fact that the rise time on LVDS is between 100 and 300pS,
> >which
> > > > > makes even 0.1" of stub significant. You are more likely to see about
> > > > > 1". Also, the parasitic capacitance of many connectors could be a
> >serious
> > > > > problem. Most folks who want to handle a lot of bandwidth have long
> >since
> > > > > gone to point to point configurations for these reasons.
> > > > >
> > > > > Good luck with your MPEG2 endeavor.
> > > > >
> > > > > Regards,
> > > > >
> > > > > Steve.
> > > > > At 10:40 AM 1/13/00 -0700, you wrote:
> > > > > >"Bruce W. Marler" <BMarler@WooshCom.com>
> > > > >
> > > > >
> > >
> > >
> >
> >
> >**** 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
> >****
>
> **** 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
> ****

--
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 ****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Thu Apr 20 2000 - 11:34:38 PDT