[SI-LIST] Re: Remove Ground underneath Differential signal isdeserved or not?
- From: yun <yunnan@xxxxxxxxxx>
- To: Scott McMorrow <scott@xxxxxxxxxxxxx>
- Date: Thu, 8 Aug 2002 17:38:39 +0800
Scott,
Thank you very much for your excellent explain!
I completely agree with you that plane separation makes high frequency signal
crossalk quite small. However, for high frequency digital signal, it¡¯s
spectrum occupy from Dc to nearly 1/(3*Tr), within this frequency range, the
amplitudes are the same for different frequency, that is what I concern. From
your description, the crosstalk from traces driven underneath the ground plane
is about -56 dB with 1/2 oz copper at 10MHz, then for lower frequency, the
crosstalk will be larger. At last, the crosstalk will be large enough and can
not be neglected, and it will influence the integrity of high frequency digital
signal. Is this thought correct?
Another thing, you have said that¡±Spice simulations showed that the crosstalk
matched the VNA measurement results within -1 dB across several different trace
configurations.¡± What are the different configurations? And is the horizon
distance effect also within -1 dB?
Best regards,
Yun Nan
Senior Engineer
Wuhan Fiberhome Networks
yunnan@xxxxxxxxxx
Yun,
The penetration of a signal through a meterial is described by an
exponential function and is:
e(-alpha * z) the exponential function of (- alpha * z)
z is the distance below the surface.
we call 1/alpha the skin depth
where
delta = 1/(sqrt (pi * f * mu * sigma))
This is the point where the field strength has dropped 37 percent. But
field still penetrates the copper and is described by the exponential
function:
e**(sqrt (pi * f * mu * sigma))* z
As the distance z increases, the field strength falls off exponentially,
and depends on frequency. The field that penetrates the plane is quite
small. And at high frequencies is neglibible. But for low frequency
analog work, the penetration may be enough to cause isolation problems.
In my case, I was trying to isolate 10 MHz Video signals in a video
switch from one another by better than -66 dB.
(-60 dB for the crosstalk from two neighbor traces.) We used guard
trace to reduce adjacent neighbor crosstalk to below -70 dB, but found
that there was still excessive crosstalk from traces driven underneath
the ground plane of about -56 dB with 1/2 oz copper.)
In analyzing the problem, I used Ansoft 2D, which is a finite element
mesh field solver, where the mesh size and error energy for the
extraction can be user controlled. (At much pain, I might add).
By reducing the mesh size, using a fine mesh on the plane and reducing
the error energy to -70 dB, I was able to get an extraction of an
equivalent circuit which predicted crosstalk betweeen traces separated
by a plane. Spice simulations showed that the crosstalk matched the VNA
measurement results within -1 dB across several different trace
configurations.
Since our plane thickness was 1/2 oz, we were able to increase it to 1
oz and saw nearly a -10 dB improvement in crosstalk isolation.
This is not an important effect for high frequency digital signalling.
However, it might be for lower frequency analog signals, such as audio
and video, and is the reason why audio planes are often will isolated
from all others. In my case, the total video signal trace length
through the switch was about 70 inches. It didn't take much to develop
-66 dB of crosstalk.
You can find a treatment of skin depth and field penetration in most any
undergraduate electromagnetic fields textbook.
regards,
scott
--
Scott McMorrow
Teraspeed Consulting Group LLC
2926 SE Yamhill St.
Portland, OR 97214
(503) 239-5536
http://www.teraspeed.com
yunnan wrote:
>> Scott,
>> For what you say: ¡°In the same project I also was successfully able to
>> extract with a 2D field solver, simulate with Hspice and correlate to
>> measurements, the crosstalk isolation between two traces seperated by a 1/2
>> oz copper plane. Through simulation we were able to determine that by using
>> 1 oz copper we could increase the crosstalk isolation by 10 dB.¡± you have
>> mentioned the crosstalk between two traces seperated by a plane, can you
>> make a more detail description? I am very interested in it.
>> Best regards,
>> Yun Nan
>> Senior Engineer
>> Wuhan Fiberhome Networks
>> yunnan@xxxxxxxxxx
>> -------------------------------------------------------------------------------
>> I've been able to successifully achieve better than -70 dB of crosstalk
>> isolation between multiple analog video signals using guard traces for
>> over 60 inches of total trace length. They do work quite nicely if vias
>> are used to ground the trace over regular intervals.
>>
>> Not only did I have these results in Spice simulations. But also they
>> were subsequently confirmed by VNA measurements on the traces, within 1
>> dB of the simulated results. Without the guard traces, the crosstalk
>> was greater than 6 dB worse. In the same project I also was
>> successfully able to extract with a 2D field solver, simulate with
>> Hspice and correlate to measurements, the crosstalk isolation between
>> two traces seperated by a 1/2 oz copper plane. Through simulation we
>> were able to determine that by using 1 oz copper we could increase the
>> crosstalk isolation by 10 dB.
>>
>> We utilized guard traces for several hundred single-ended and several
>> hundred differential traces on an analog video crosspoint switch.
>> Without the guard traces and stitch vias, we would not have been able
>> to meet our end-to-end crosstalk isolation requirement of -60 dB.
>>
>> Best regards,
>>
>> Scott
>>
>>
>>
------------------------------------------------------------------
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 archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
