Re: SI- Termination Comments Wa

Larry Smith (ldsmith@faithful-188)
Mon, 22 Apr 1996 09:05:20 +0800

Norman - there is a real convenient 'rule of thumb' that Paul Franzon taught
me in one of his signal integrity short courses. It has to do with the
transmission line being 1/3 of a rise time or less. If the electrical
discontinuity (transmission line section, stub, electronic package, connector,
anything-you-can-think-of) is less than a third of the rise time, then the
near end of the discontinuity has a chance to influence the far end of the
discontinuity during the rise time. The signal goes down, back and down again
during the rise time. This keeps the far end from getting too far away (in
voltage) from the near end, and the structure can be assumed to be a lump rather
than a more complicated distributed network. If the discontinuity is in a signal
that you really care about (ie clock), then use the 1/5 rise time rule so that
the near end has a chance to influence the far end twice during the rise time.

For your example, assume that signals travel at 6 inches per nSec in FR4.

5 nSec rise time:
1/3 rise time = 5/3 = 1.66 nSec
distance in FR4 = 6 in/nSec * 1.66 nSec = 10 inches
1 nSec rise time:
1/3 rise time = 1/3 = .333 nSec
distance in FR4 = 6 in/nSec * .333 nSec = 2 inches

Therefore, with a 5 nSec rise time, a transmission line section or stub that
is less than 10 inches can be considered to be a lump. For a 1 nSec rise time
(which is more typical for today's components) any FR4 structure longer than
2 inches must be analyzed as a transmission line. Conversely, stubs less than
2 inches just look like extra lumped capacitance on the daisy chain net.

On the termination question, Yes, you need to terminate. There are lots of ways
to terminate. For nets with just one driver, I like the series resistor between
the driver and one end of a daisy chain net. Values around 40 Ohms work well.
For nets with lots of drivers (ie data bus) - good luck! I have had success
with a near end cluster separated from a far end cluster by a series resistor.
Keep the point to point distance in the clusters less than 1/3 of a rise time
(guess why...).

Of course, the best way to evaluate this is with Spice simulation with all
of the packaging carefully modeled. After doing this several times, you will
finally come down to some fairly simple rules of thumb as shown above.


----- Begin Included Message -----

I am currently working on a mid-speed (<50MHz) digital circuit pack and would like to proper terminate some long lines(up to 15 inches). It consists of CPU, DSP, SRAM, Flash etc. I am approaching the problem by using Daisy-Chained tracks and AC terminations. There are some findings that I would like your comments:

1. When using uncontrolled-impedance FR4 PCB, based on my calculation on a 6 layer board, the micro-strip (8 mil) Z0 is about 75 Ohm and the micro-strip Z0 is about 45 Ohm. That means a mismatch every time I switch layers. Does anyone has experience on this? Does it matter for 5 ns rise time? How about 1ns rise time?

2. My EDA simulation package showed that after termination, my signals do not look much cleaner. In fact, it look worse and seems to be loaded down (Vpeak is about 3.5-4V instead of 5V). Does this make sense?

3. At want point could I use T instead of Daisy-Chain so that the stub look like capacitance, no transmission lines? A lot of time daisy chain line is longer than a treed line.

Norman Wong
Hardware Design Engineer
Nortel, Wireless Development Center, Calgary

----- End Included Message -----