[SI-LIST] Re: High speed signal on top layer
- From: "Joe Paul M" <joepaul@xxxxxxxxxxx>
- To: "si-list" <si-list@xxxxxxxxxxxxx>
- Date: Tue, 31 May 2005 12:54:43 +0530
I don't think that referring a differential signal to power plane has
any serious issues, provided the plane is continuous from driver to
receiver.
I understand that most systems have a good continuous GND plane
maintained with great care, unlike power plane, which may be the reason
why people prefer GND plane as reference plane.
Your cards (daughter cards) which plug in to backplane will mostly have
the signals referred to GND plane. When these signals enter backplane,
there will be a discontinuity in case you refer the signals to power
plane in the backplane.=20
If you can workout a plan to have the signals refer to same power plane
in your daughter cards also, you won't have much problem in using power
plane as ref plane. (At your speed, stitching/decoupling one ref plane
to another becomes very difficult/ineffective, due to the parasitic
inductances associated with mounting, via etc)
Regards
Joe Paul M
=20
=20
=20
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of Larry Miller
Sent: Tuesday, May 31, 2005 9:57 AM
To: george_dai@xxxxxxx
Cc: si-list
Subject: [SI-LIST] Re: High speed signal on top layer
George,
On May 30, 2005, at 8:37 PM, George Dai wrote:
> Hi Larry,
>
> Thanks for your comment,
> 1, Currently we limit the delta of the pos/minus signal lentgh to less
> than 50 mil, is that enough? (Per my caculation, it create around 10ps
> different delay)
Should be OK.
> 2, I don't understand the EMI issue very clearly. Do you mean the 3G
> signals will couple to the adjacent power plane, then make EMI? Any
> other pitfalls besides EMI problem?
The signals need to return to the driver via a GND plane. (Most =20
"differential pairs" are not really differential but are single-ended =20
signals driven out of phase. For it to be truly differential you =20
would need transformers, like in Ethernet 100BASE-TX)
If a good GND is opposite your signal traces there is no problem. The =20
signal currents mirror onto the GND plane and return to the driver.
If your signals are opposite a power plane, then the return currents =20
mirror onto that plane. However, these currents have to ultimately =20
get to a GND plane to complete the circuit (Kirchoff's Law). If you =20
have appreciable impedance between the power plane and the GND plane, =20
then two things will occur:
1) The signal current IR drop (or, really, IZ since it is usually =20
complex) will make noise on the power plane. This can get radiated =20
(EMI).
2) The added impedance between power and GND planes add to the =20
overall transmission line systems impedance, causing an impedance =20
mismatch, which in turn causes reflections on your differential =20
pairs. (SI problems, maybe more noise and EMI.)
At lower frequencies (below 100-200 MHz) you can effectively bypass =20
power to GND, i.e., create an AC short circuit, with capacitors. =20
Above this, discrete capacitors do not work well (See si-list for =20
discussions). Up to 500 or 600 MHz you can create "good" bypass =20
capacitors by having power and GND planes closely spaced, so you can =20
maintain a low AC impedance from power to GND.
After you get much above 500 or 600 MHz (in frequency, or, equivalent =20
rise times) even the power/GND capacitance does not work well. This =20
is why higher frequency signals use these "differential" pairs-- the =20
signal currents largely cancel IF they have a "good" DC plane to work =20
against. Unfortunately, a "good" DC plane means that it has a low =20
impedance path back to the driver, and most ICs require that this be =20
a GND plane. This (use a GND plane) is certainly the "safe" way to do =20
it.
But there are exceptions:
As you get up above 2 GHz, FR-4 PC board material gets so lossy that =20
the return currents get dissipated rather than returned cleanly to =20
the IC. If you look at the Fibre Channel physical layer standard (FC-=20
PI-2 at t11.org), you will see that at 4 GHz they do not require =20
impedance matching at all! (0 dB S11 return loss.) So if you are =20
using a good run-length-limited coded serial scheme such as 8B/10B =20
coding, all of your frequencies will be very high, and you might have =20
a chance using the power plane for your differential pairs. (You may =20
have to do a PC board "spin" to verify this.) Most EMI problems these =20
days seem to be in the 500 MHz - 1 GHz zone, in my observation.
If possible, use Current-Mode Logic (CML) for your 3 GHz signals. It =20
seems to be much better behaved and tolerant than PECL or LVDS. My =20
experience, anyway.
Larry Miller
>
> Best Regards,
> George
> ---original message---
> From:Larry Miller ; Mon, 30 May 2005 10:15:19 -0700
> Subject:[SI-LIST] Re: High speed signal on top layer
>
>>
>> On May 30, 2005, at 4:48 AM, George Dai wrote:
>>
>>
>>> Hi all,
>>>
>>> We are layout one 3.0Gbps high speed backplane. The original design
>>> style
>>> is routing the edge coupling differential signals in the internal
>>> layer,
>>> with adjacent GND plane. Now we met the layer limitation, and here
>>> is the
>>> questions,
>>> 1. Can we routing the differential pair on the top layer with GND
>>> plane
>>> on the second layer? In my knowledge, it seems functional but odd.
>>> Please give your advice.
>>>
>>
>> That will work, but you have to be very careful to keep your routes
>> exactly the same length and parallel. If there is any unbalance it
>> radiates. Be sure to take into account the difference in propagation
>> velocity compared to striplines if you have critical timing.
>>
>>
>>> 2. Can we use one continuous power plane (12v or 5v) as the =20
>>> reference
>>> plane to the high speed signals? For example, if 3.0Gbps signals on
>>> top
>>> layer works, can the second layer be +12v power plane?
>>>
>>
>> I would not do this. It is practically impossible to make the power
>> plane at AC ground (by bypassing or capacitance between planes) at
>> 3G. You would probably have EMI problems.
>>
>> Larry Miller
------------------------------------------------------------------
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 FAQ wiki page is located at:
http://si-list.org/wiki/wiki.pl?Si-List_FAQ
List technical documents are available at:
http://www.si-list.org
List archives are viewable at: =20
//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
=20
------------------------------------------------------------------
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 FAQ wiki page is located at:
http://si-list.org/wiki/wiki.pl?Si-List_FAQ
List technical documents are available at:
http://www.si-list.org
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:
