[SI-LIST] : RE: [SI-LIST]: Long bus or star?

Grebenkemper, John (john.grebenkemper@compaq.com)
Thu, 01 Apr 1999 10:31:09 -0800

There have been a number of messages recently on using a star
configuration
for a multi-drop net. We have successfully used the star terminated
nets in
several designs. A paper I presented at DesignCon98 discusses star
terminations (a PDF copy is available on request).

The star termination can be implemented by using a resistor located
between
the junction of the star and each transmission line. This requires N
equal
valued resistors for a N net star configuration. To obtain a perfect
match
at the star, the value of each resistor needs to be set to (N-2)Zo/N.
This
will avoid all reflections at the star, but will not necessarily yield
the
best waveform at the loads. Some tuning of these values can improve the
waveform.

For instance, at N=3 each resistor needs to be set to Zo/3 for a perfect
match. For an ideal voltage source driver, this will give a maximum
overshoot of 50%. Increasing the resistor value to 0.6Zo decreases the
maximum overshoot to 18%. For real drivers, this value would need to be
somewhat less due to the effective series resistance of the driver.
SPICE
simulations can allow one to tune to an optimal resistor value in one's
system.

This resistor value can be optimized for a larger number of loads in the
star. Note that it is difficult to get incident wave switching for N>5.
Equal length nets from the star can improve the waveforms due to some
cancellations that occur in the reflections. However, quite acceptable
peformance can be obtained with arbitary length transmission lines on
the
arms of the star.

We have found that the propagation time on the net can be considerably
less
for a star configuration than for a daisy chain, especially when the
loads
off the daisy chain require longer transmission lines due to routing
constraints. The star may also work better in configurations in which
any
receiver can become the driver since each driver now sees a more
identical
load.

-John Grebenkemper
john.grebenkemper@compaq.com

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****