Keeping an eye on the original question, which was the choice between: > Option1: > L4 - Power > L5 - Signal > L6 ? GND > > Option2: > L4 - GND > L5 - Signal > L6 ? GND, ...are we agreed that these options are equivalent from a high frequency standpoint? And that the potential problems you mention regarding layout apply equally to power and ground plane implementations, and therefore are a separable problem from the choices shown above? Current always needs a return path. For a high frequency signal sandwiched between the planes of either option, the important concept is that the return current flows on both planes, in a ratio determined by the proximity to each. For a trace centered between them, the return current in each plane will be half that on the signal trace. Of course, any DC component is restricted to returning via ground, but that is not the same as the AC case. There is a classic EMC demo that shows the transition region between the two regimes, but basically, once the frequency is high enough for the bypass caps to make a difference, then both planes in the microstrip sandwich participate as return paths, and it doesn't matter whether both are ground planes, or one is power and one is ground, or both are power, all else being equal. In other words, the assignment of a plane to power or ground is not, in and of itself, an SI issue. BTW, it is not entirely true that differential pairs have no ground return current. If a plane is close enough to affect the impedance of the pair, it will carry ground currents in both directions at opposite polarities, one for each trace. There will be a current null at the point midway between the traces and a peak underneath each trace. The sum will be zero, but the actual currents are not each zero because the simple fact that the signals are balanced does not mean that the distance between the traces has no effect. For sufficiently distant planes or small enough separations (or both), the paired return currents might be trivially small. For typical board stacks and trace pitches, Murphy says they're not. Orin On Sun, 23 Sep 2007 10:11:03 -0700 (PDT) Jory McKinley <jory_mckinley@xxxxxxxxx> writes: This is interesting and hopefully not too long winded and off the track........... Correct me if I have this wrong but current always attempts to return to its source no matter what frequency we are talking about, if it does not then we have a problem such as radiation in the path. For AC signals the return or image current is instantaneous to the traveling signal attempting to get back to the source through any low inductance path. This path may be the opposite line for differential it may be as you pointed out power or ground or both. However, there could be many hurdles for this return current, signal via's, reference plane splits, and as you mentioned improper decoupling to name a few. One of the issues on a board with a gnd/sig/gnd configuration is the potential long return loop for single ended signals routed on this layer and potential issues with any unbalanced common mode differential noise. I will say that for "ideal" differential with well balanced decoupling, and return vias the gnd/sig/gnd works but ideal is not that easy. -Jory ----- Original Message ---- From: "olaney@xxxxxxxx" <olaney@xxxxxxxx> To: jory_mckinley@xxxxxxxxx Cc: raja.anand@xxxxxxxxx; si-list@xxxxxxxxxxxxx Sent: Sunday, September 23, 2007 11:17:49 AM Subject: Re: [SI-LIST] Re: High speed signal-return path That makes sense for DC current, but the point is that the power and ground planes are shorted together in the RF sense by a properly implemented bypass network. Therefore, one is as good as the other for RF. RF energy never reaches the power supply -- if it does, something is very wrong. Orin On Sat, 22 Sep 2007 22:11:06 -0700 (PDT) Jory McKinley <jory_mckinley@xxxxxxxxx> writes: Hello Orin, There is in that the return current has to do just that, return. So for single ended or common mode differential sourcing or sinking from power or ground could make a significant difference, somewhat regardless of edge rate. All mute for ideal differential of coarse. -Jory ----- Original Message ---- From: "olaney@xxxxxxxx" <olaney@xxxxxxxx> To: jory_mckinley@xxxxxxxxx Cc: raja.anand@xxxxxxxxx; si-list@xxxxxxxxxxxxx Sent: Saturday, September 22, 2007 12:49:27 PM Subject: [SI-LIST] Re: High speed signal-return path We all know differential versus single ended, but "return current plane, not ground" needs explanation if you feel that there is some nontrivial difference at high frequencies between power and ground planes. Orin On Sat, 22 Sep 2007 10:36:14 -0700 (PDT) Jory McKinley <jory_mckinley@xxxxxxxxx> writes: Yes, however every plane is a potential return current plane, not ground, less important for true differential more important for single ended. How about any common mode noise on the differential signals, I would think Option 1 would allow for better potential isolation. -Jory ----- Original Message ---- From: "olaney@xxxxxxxx" <olaney@xxxxxxxx> To: raja.anand@xxxxxxxxx Cc: si-list@xxxxxxxxxxxxx Sent: Saturday, September 22, 2007 6:51:30 AM Subject: [SI-LIST] Re: High speed signal-return path Every power plane is a ground plane for high frequency purposes. There should be no material difference between the two implementations. If there is, something is really wrong with the power supply bypassing strategy. The choice between stripline and microstrip is generally driven by the particulars of the board layout. There is no definitive choice of one over the other based on general principles, only as circumstances dictate. Orin Laney On Sat, 22 Sep 2007 11:12:19 +0530 "Raja Anand" <raja.anand@xxxxxxxxx> writes: > Hi All, > For high speed differential signals (3.125Gbps ), is it really > required to > take these signals between Gnd Planes. Which option should I go for > from > below if my stack up is symmetrical? > > Option1: > L4 - Power > L5 - Signal > L6 ? GND > > Option2: > L4 - GND > L5 - Signal > L6 ? GND > > Regards, > Raja > > ------------------------------------------------------------------ > 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 technical documents are available at: > http://www.si-list.net > > 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 > > > > ------------------------------------------------------------------ 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 technical documents are available at: http://www.si-list.net 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 Check out the hottest 2008 models today at Yahoo! Autos. ------------------------------------------------------------------ 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 technical documents are available at: http://www.si-list.net 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 Take the Internet to Go: Yahoo!Go puts the Internet in your pocket: mail, news, photos & more. Be a better Globetrotter. Get better travel answers from someone who knows. Yahoo! Answers - Check it out. ------------------------------------------------------------------ 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 technical documents are available at: http://www.si-list.net 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