Re: [SI-LIST] : Signal Ground at the connector

Roland F. Portman (portman@bitmicro.com)
Thu, 17 Dec 1998 12:39:48 -0800

Eric,

With a 2ns tr/tf data signal you have a 500MHz effective switch rate. For this level
of signal you will need to match the impedance of the each board, the
connectors and the cable to reduce the reflections, crosstalk and radiated
energy. You may even need to shield the cable in addition.

The old standard grey flay ribbon cable is a poor medium for transmitting
such a fast signal. The impedance of the cable varies
widely from cable vendor to cable vendor which mainly has to do with the
different materials (dielectric constant) they use to make the ribbon cable
and the amount of material surrounding the conductors. This can vary
from batch to batch as well. Be sure to pick a 'good' cable vendor.

You can do a few things which will help (in order of my preference).

1. Interspace grounds between every signal G-S-G-S-G-S-G-S-G-S-G-S-G.
This will help with crosstalk, but may not be sufficient for matching the
impedance of your board.

2. Slow the rise time of your signals. If you have single ended (one source to one recevier)
then you can use series terminators. Be sure the resistor is less than 0.5" from the output pin.
CTS and others make very small SMD isolated resistor arrays which are ideal for multiple series terminations.

3. Don't use the old gray flat IDC ribbon cable. There are newer products which are
designed for higher speed cables. There is the finer spaced 2mm cable.
There are teflon based ribbon cables with finer spacing as used for Ultra-SCSI.
(This cable was designed for 40 MHz signals to travel 3 meters.)
And there are much better connectors than IDC, such as AMP Mictor for example.
Do a web search for +cable +controlled +impedance.
And look at 3M, AMP, Woven, Berg, RobinsonNugent just to name a few.

4. Lastly, use the filters. I haven't the experience to address this properly. I would
guess if you are filtering the signal then you would ground to signal-ground. If you are
filtering the noise on the cable, then you would ground to chassis-ground. Be very
cautious of using moats.

I hope this gives you a start. Others will be responding as well.

Roland

ycchien wrote:

> Hi,
> I'm currently designing a circuit board with ribbon cable running
> outward to another daughter board. The signals' tr/tf time riding on the
> cable are approximately 2ns. To eliminate the RF energy enemating from
> the signal line, EMI filter is considered to be inserted along the
> signal line near the connector before it leaves the board.
> After reading some articles on filtering the signal line, I'm still
> confused at how to ground the filter. Most of the articles suggest
> grounding the filter to the chassis ground not the signal ground and
> then capacitively couple these two grounds. Several measures can be
> adopted, they are
>
> A. Introduce a chassis plane to the layer stack, the filter ground is
> connected through the via to the this chassis plane. But, where should
> the signal return grounds of the cable be connected to? The chassis
> plane or the signal ground plane?
>
> B. Moat the signal ground around the connector, leaving a wide piece of
> copper to serve as bridge accross the moat. Reserve srew holes on the
> ground plane of the connector part to provide a shorted path to the
> external chassis plate, when low frequency isolation is not a concern.
> However if low frequency isolation do matter, how to capacitively couple
> this ground plate to external chassis plate?
>
> C. Moat the signal ground along the board circumference, with the
> segmented ground plate along the circumference to serve as chassis
> plane. The filter and signal return ground are connected to this plane.
> The same question is encounterd as in item B.
>
> Any comments and advices are greatly appreciated?
>
> Eric Chien
> Assistant Researcher
> Synchrotron Radiation Research Center
> Taiwan, R.O.C.
> ycchien@alpha1.srrc.gov.tw
>
> **** 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 ****

--
Roland F. Portman
BiT Microsystems, Inc.
48499 Milmont Drive
Fremont, CA  94538

(510) 623-2341 Ext 129 (510) 623-2342 Fax

portman@bitmicro.com http://www.bitmicro.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/si-list ****