I do not believe that you should use a shield layer connected at a
single point to ground (ground in this case is earth (green wire in
power cord) and logic ground). I am assuming that your system's clocks
are in the range of 50mhz - 200mhz ( i.e. your system does not run in
the KiloHertz range). Here is my thinking:
The basic idea of having a logic ground and shield ground connected at
a single point (generally at the power supply) is to create inner and
outter Faraday cages. In the ideal case the outter F-cage shields the
inner F-cage because of the high impedance single-point
connection. Obviously if the F-cages were ideal there would be no need
for an inner and outter cage. Since they are not ideal (i.e. seams,
holes, slots, cables, etc.)
(1) Currents flowing in the shield layer due to accidental coupling
from the logic will have an electrically-long, high-impedance return
path to the source. This increases radiation. General rule is that
return paths should be electrically short (minimizes
coupling/radiation below ideal antenna length) and low impedance
(minimizes electric field potential).
(2) Cable shields. Currents flowing in the cable shield as a result of
single-ended signaling will need a wide-bandwidth, low-impedance
connection to logic ground. Creating such a connection while
attempting to maintain isolation between shield and logic ground at
the connector interface is difficult mainly because of the upper
frequency limit versus C-value of most caps.
In circuit board construction you can create the shield layer directly
over the logic ground layer so that they are closely coupled
(i.e. low-impedance), but then I would wonder why you don't just do
the logic ground layer correctly in the first place.
I use the single F-cage approach with a unified multipoint grounding
scheme sucessfully on large multi-cabinet systems. Please feel free to
email me on some of the other issues, at your convience.
For further reading on the subject, please see "Noise Reduction
Techniques in Electronic Systems", Second Edition by Henry Ott, ISBN
0-471-85068-3.
*************************
Thank you |
Don Abernathey |
(503)690-6234 |
[email protected] |
*************************
On Jan 8, 5:48pm, Joachim Mueller wrote:
> Subject: shield layer connection
> hi,
>
> I have had a discussion with a colleague from SNI about the location of the
> ground connection of shield layers in multiple PCB (computer-)systems.
> The ideal shield-layer is a currentless plane on the component sides of
> a multi layer PCB.
>
> Till today we connect the shield layers of the PCBs only to the
> shield layer of the backplane which is connected itself at one single point
> with system ground.
> With respect to the aim of better decoupling between different PCBs we
believe
> that it might be better to connect the shield layer of each PCB at the
board
> connector with some groundpins and the shield of the backplane.
>
> Are there any experience or comments to one ore both of this shield
> techniques?
>
>
> A happy new year to everybody!
>
> Regards,
>
> Joachim Mueller
>
> --
> +---------------------------------------------------------------------+
> | voice: +49 +5251 606-178 | Joachim Mueller, Cadlab |
> | fax : +49 +5251 606-155 |Fuerstenallee 11, D-33102 Paderborn,|
> | e-mail: [email protected] | Germany |
> +---------------------------------------------------------------------+
> | cadlab/Analog System Engineering |
> | - Joint venture University of Paderborn and |
> | Siemens Nixdorf Informationssysteme AG |
> | |
> | __o |
> | _`\<,_ |
> | (_)/ (_) |
> | ~~~~~~~~~~ |
> +---------------------------------------------------------------------+
>-- End of excerpt from Joachim Mueller