Re: [SI-LIST] : PCB track length and ringing

Larry Smith ([email protected])
Thu, 25 Sep 1997 08:53:05 -0700

Dmitri makes a real good point below. I use the '1/3 rise time' rule of
thumb. If the electrical length of the trace (in nSec) is less than 1/3
the rise time, the trace can be considered to be a node. No
transmission line analysis is necessary unless you are really concerned
about this net (clock trace or something). If the trace is more than
1/3 of a rise time long, reflections (ringing) will occur if the
transmission line is not terminated.

For a 1 nSec rise time, a trace longer than .33 nSec (2 inches in
FR4) must be considered to be a transmission line.

There is a physical interpretation to this rule of thumb. If the
wavefront has time to go down the trace, back and down again (3 times of
flight) during the rise time, then the near end can influence the far
end during the rise time. If the trace is longer than that, the far end
goes out-of-control. For a typical CMOS receiver (open circuit) the
wavefront will double at the far end and ringing will result.

Larry Smith
Sun Microsystems

> I would like to point out the effect the switching speed and
> interconnect length have on the ringing.
> The magnitude of ringing cannot be larger than the transient slew rate
> times the round-trip delay thru the trace. For the ringing to reach its
> maximum magnitude under given terminations, the interconnect round-trip
> delay has to be greater than the switching time.
> As a result, even for large reflections and small loss, the ringing may
> be almost unnoticeable in slower circuits and shorter interconnects, and
> becomes more pronounced as switching speed and interconnect length
> increase.
> Regards,
> -- Dmitri