kicad-developers team mailing list archive

[Andrew Zonenberg <azonenberg@xxxxxxxxxxxxxxx>]

Length matching in USB or other differential signals is for avoiding
phase offsets between the two halves of the differential pair, length
matching in RAM or other parallel data buses is reduction of skew to
ensure sufficient setup time between DQS/clock and data arrival.

In both cases the issue is not length matching per se, it's delay
matching. If you assume the board is uniform composition and impedance
over its area (which is valid if there are no breaks in the planes and
the FR4 substrate is good quality), all signals in the bus are on the
same layer, and there are no discontinuities like vias in the signals,
then delay matching approximately reduces to the simpler case of length
matching.

Propagation delay is typically on the order of 6.5 ps/mm but is higher
on buried lines than surface lines and is dependent on a bunch of other
factors too.

Micron gives a good example of static timing analysis for DDR3 1600 in
TN4113 on page 11-12. For a 476 ps window, after subtracting all of the
other sources of eye narrowing (controller skew, crosstalk, rise/fall
times, etc) there's only 10ps of slack allocated to routing delay which
means length matching of +/- ~0.6mm across the entire bus.

On Wed, 2014-04-16 at 22:40 +0200, Pierre Parent wrote:
> Hi,
> 
> Well, someone reported to me that he could ran a board only in slower DDR 
> timing because he did not do any  length matching.
> 
> Also olimex has done length matching in their board olinuxino maxi (enclosed 
> image), and we can find plenty of design guidelines about DDR(2,3) routing 
> speaking of length matching. Eagle and Altium include length matching tool. So 
> I'm not an expert but it is probably not just a joke when it comes to Ram 
> (don't know about USB).
> 
> Pierre.
> 
> 
> 
> Le mercredi 16 avril 2014, 22:01:06 Lorenzo Marcantonio a écrit :
> > On Wed, Apr 16, 2014 at 08:33:45PM +0200, Pierre Parent wrote:
> > > I've sent to production my first length matched board, here is a picture!
> > > I
> > > hope it will work!
> > 
> > I did a board which contained an USB high speed hub (and an USB camera,
> > too). Didn't measure the impedance or looked at the eye pattern (no
> > equipment :P) but it worked. Luckily both TVS and common mode chokes
> > where designed to work with it, so no big meandering to do...
> > 
> > So it should work your too.
> _______________________________________________
> Mailing list: https://launchpad.net/~kicad-developers
> Post to     : kicad-developers@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~kicad-developers
> More help   : https://help.launchpad.net/ListHelp

-- 
Andrew Zonenberg
PhD student, security group
Computer Science Department
Rensselaer Polytechnic Institute
http://colossus.cs.rpi.edu/~azonenberg/

Attachment: signature.asc
Description: This is a digitally signed message part