I made a test board which has a 5mil 50ohm trace, launched with an SMA connector, 2" to an 3/4" S bend, 2" to a via, 2" back to an identical 3/4" "S" bend but with square corners, and 2" back to an 51ohm 0603 terminator resistor under the SMA connector. The via is a 10mil finished drill size with a 20mil pad, and a 34mil antipad to the planes. The stackup was Top Trace 3 mil FR4 Ground Plane 1.8 mil FR4 Power Plane 36 mil FR4 Power Plane 1.8 mil FR4 Ground Plane 3 mil FR4 Bottom Trace There was a ground shorting via between the two ground planes 75 mils from the thru via at the end of the trace. When looking at it with the TDR, you can clearly see both S bends, the via, and the terminator resistor. Both S bends are basically the same, rounded corners vs square (a total of 6 corners each) made no noticeable difference. The via was clearly inductive, the impedance went up about 6 ohms, and the pad on the terminator resistor was capacitive with an impedance dip of about 5 ohms. I mention this in reference to the comment of untuned via's generally being capacitive in nature. I have seen several reports of the effects of vias, but no one ever puts actual numbers down as to the impedance and geometries. My numbers might be slightly off, they are from memory, but they are close. Wilbur Harvey - Engineer Adtech Inc., www.adtech-inc.com 3465 Waialae Ave. Suite 200 Honolulu, HI 96816 Tel: +1 (808) 440 3363, Fax: +1 (808) 440 3494 email: wilbur.harvey@xxxxxxxxxxxxxx -----Original Message----- From: pikeda@xxxxxxxxxxx [mailto:pikeda@xxxxxxxxxxx] Sent: Thursday, February 14, 2002 6:09 PM To: si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Diff line Scott, I agree with everything you said and would like to add that loss due to skin effect will increase since the extra capacitance in the lines will tend to squeeze the flow towards the inside edges. Fabrizio, Pravin, if the multigig lines are that dense that you can't avoid tightly coupled lines than I would suggest matching lengths near transitions such as at the via, bga or connector. After all if you look at an untuned vias, they generally look capacitive or dip in impedance. The seperated line is higher in impedance. So if the risetime is long compared to those two structures that are next to each other, than the impedance should actually look more flat. For 2.5GHz I estimate that the 2 structures should be less than 300 mils (very rough estimate). Another tip if EMI is not too much of a concern such as for well referenced stripline, is to match some distance away from the driver where the losses have slowed down the risetime and the mismatch in impedance will have less of an impact. No alternative solution is without a downside for certain circumstances or it wouldn't be an alternative, it would just be the only solution. Paul "zanella, fabrizio" <zanella_fabrizio@xxxxxxx> on 02/14/2002 06:39:44 PM Please respond to zanella_fabrizio@xxxxxxx To: Paul Ikeda/Marvell@xxxxxxxxxxx, si-list@xxxxxxxxxxxxx cc: Subject: [SI-LIST] Re: Diff line Paul, a large percentage of nets on pc boards are now differential. By using 20 mil spacing for your loosely coupled differential pairs, I don't see how you can route high density boards without using 20-30 layers, which increases cost, creates manufacturability problems, increases via capacitance, etc. The trend is to pack more and more routes in a given space. 5 mil lines with 5 mil spacing, edge coupled on stripline layers is quite common for differential pairs and works very well for 2.5Gbs signals. For the differential skew value, make sure you look at the entire path, through the BGA, connectors, vias, etc, the total lenghts may be closer than 300 mils. Regards, Fabrizio Zanella Signal Integrity EMC Corporation fzanella@xxxxxxx -----Original Message----- From: pikeda@xxxxxxxxxxx [mailto:pikeda@xxxxxxxxxxx] Sent: Thursday, February 14, 2002 9:08 PM To: si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Diff line Pravin, This is just one more argument to loosely couple your differential lines. If they have around 20 mils seperation, you can treat them as single ended and serpentining has no effect on impedance. Here is an article by Dr. Howard Johnson on why you don't need to tightly couple your diff pairs which I agree with wholeheartedly. By the way I commonly do 3.125GHz loosely coupled and they work very well. http://www.sigcon.com/news/2_30.htm Paul " pravin patel" <fairfax100@xxxxxxxxxxxxx> on 02/14/2002 06:25:04 AM Please respond to fairfax100@xxxxxxxxxxxxx To: si-list@xxxxxxxxxxxxx cc: (bcc: Paul Ikeda/Marvell) Subject: [SI-LIST] Diff line I am doing layout of 2.5Ghz diff. line and having hard time to match length of a pair by 300 mil. I was going to make one line serpentine to make both line matched. By doing serpentine, I am creating impedance discontinuty or I will create skew. Any Ideas. Thanks -- ------------------------------------------------------------------ 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 ------------------------------------------------------------------ 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 ------------------------------------------------------------------ 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 ------------------------------------------------------------------ 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 ------------------------------------------------------------------ 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