[SI-LIST] Re: Adding trace length for timing adjustment
- From: Tegan Campbell <Tegan.Campbell@xxxxxxxxxxx>
- To: si-list@xxxxxxxxxxxxx
- Date: Tue, 20 May 2003 11:28:20 -0600
Just a thought.....I agree from a technical perspective if you are talking
about 10's of picoseconds matching. But on a board I just finished I
instructed the designer to run three clock lengths on inner layers, and
match them to within 50ps in simulation. He used serpentining with a rule
that the net could not come within 2*dielectric thickness of itself. A
fourth clock net I had him run on the outer layer and have it's flight time
be 500ps less than the others. This did two things:
1-Satisfied the customer(internal to the company, in this case) that I
thought about proper timing on the board level and conformed to his specs
for his chip
2-Impressed upon people the value of using SI tools along with normal board
flows. Saved ~2.00 per board(precision delay IC). It helped that we had no
EMI problems when we got into the chamber.
for what it's worth.
Tegan
-----Original Message-----
From: Loyer, Jeff [mailto:jeff.loyer@xxxxxxxxx]
Sent: Monday, May 19, 2003 11:23 AM
To: Jon Powell; doug@xxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Adding trace length for timing adjustment
I agree. The reference to pS came out wrong - today I would state it =
differently (I wrote the original some time ago). Thanks for pointing =
out the mis-statement.
When we get up to high enough speeds that pS's count, other techniques =
are called for. PCI-express, for instance, doesn't have the length =
matching constraints of past busses (although the matching between =
halves of differential pairs is specified).
Jeff Loyer
-----Original Message-----
From: Jon Powell [mailto:jonpowell@xxxxxxxxxxxx]
Sent: Monday, May 19, 2003 8:59 AM
To: Loyer, Jeff; doug@xxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: RE: [SI-LIST] Re: Adding trace length for timing adjustment
In my opinion if you are using serpentine nets to provide pico second
matching, you are fooling yourself and just trying to match some spec =
that
is probably suspicious in itself. The coupling of nearby traces and just
plain irregularities on the constructed board are going to wipe out any =
ps
precision matching that you think you are doing.
regards
jon
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Loyer, Jeff
Sent: Friday, May 16, 2003 6:54 PM
To: doug@xxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Adding trace length for timing adjustment
First, a question of mine (for anybody): is there some theory that =3D
declares that the effect I describe below (coupling across serpentine =
=3D
legs causes part of the wave to bypass the serpentine) is zero in =3D
stripline?
Now that I've asked that, here's some of what I've found:
________________________________
No intense research here, but I did TDR the Front-Side Bus of a product =
=3D
board that had a variety of serpentine types (a few loooong legs, many =
=3D
short legs, some in-between), and here's what I found.
Conclusions:
1. Serpentining was >80% effective on this Front Side Bus (FSB). I.E.,
=
=3D
we achieved at least 80% of the expected delay from serpentines.
2. Vp variation due to serpentining induced about 30pS of skew on this
=
=3D
3", 50ohm bus.
3. A few long legs of serpentining were better than several short ones.
4. There won't be any "Rules of Thumb" which will easily describe the =
=3D
issue (how effective are serpentines?). When we need to worry about pS, =
=3D
simulations have to be performed on expected worst-case nets.
Other Notes
1. Only layer 1 was represented here. The effects have been less on
=3D
stripline.
2. The dielectric is 4mils thick; serpentine legs are separated by =3D
20mils center to center (same as trace-to-trace separation); trace width =
=3D
=3D3D 7mils.=3D09
Some general conclusions of mine:
Increasing trace length with serpentines (AKA "meanders") does not give =
=3D
an increase in flight-time directly proportional to the increase in =3D
trace length. Coupling across the serpentine legs causes part of the =3D
wave to bypass the serpentine (I would refer to it as a "barreling =3D
through the switchbacks" phenomena), reducing the flight-time. The =3D
speed-up effects are reproducible in simulations and seem to be only =3D
weakly tied to rise-time.=3D20
The effect can be lessened by separating serpentine legs, or routing in =
=3D
stripline. A "flat spiral" (AKA "bifilar spiral") is an option mentioned =
=3D
in some papers, but my experience with it has shown it to be terrible, =
=3D
S.I.-wise.
Take pains to avoid serpentines - they're not free.=3D20
If serpentining is necessary, keep the adjacent traces far apart - =3D
perhaps an S/H (trace separation to dielectric thickness) ratio of about =
=3D
5 to 1.=3D20
Routing in stripline reduces the effect.
Jeff Loyer
-----Original Message-----
From: Doug Brooks [mailto:doug@xxxxxxxxxx]
Sent: Thursday, May 15, 2003 10:09 AM
To: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Adding trace length for timing adjustment
I know there are several different views on this subject.
Suppose I need to add some additional length (time) to a trace and =
am=3D20
considering three different strategies:
1. a randomly meandering length
2. a "trombone-like" length down and back
3. more, shorter snake-like loops
What do people see as the different trade-offs for these three different =
=3D
strategies under differing conditions?
Doug Brooks
------------------------------------------------------------------
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: =3D20
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages=3D20
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
=3D20
------------------------------------------------------------------
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
Other related posts:
