"Current steering means having a current source (not necessarily having a very high impedance)"
I appreciate your response, but there are few difficulties
here.
First, a "current source" by definition has an infinite
impedance, to speak accurately. Therefore we must assume
that when people are talking "current-source drivers",
they mean some crude approximation of this.
Second, the Rambus signaling scheme is a 28Ohm terminated
to 1.8V. If we generously assume that the Rambus
"impedance of hundreds of Ohms" is even 100 Ohm,
an open-drain N-MOS structure could not drag the wire down
by more than 0.35V, which would not enough even to cross
the reference level of 1.4V, not speaking of the
specified level of 1V. As a matter of fact, the Rambus
signaling scheme specifies 30mA as a constant driver
sink, so it must be about only 30 Ohm to drive a 28-Ohm
transmission line to 1V from 1.8V. Actually, it seems that
the driver must have a variable impedance from 60 Ohm
(at 1.7V) down to 30Ohm during the switching phase.
Also, why would you assume that the author of that
message was using a legally-correct terminology while
citing the 12-year-old SCI presentations from memory?
After all, it could be very beneficial for Rambus to
adopt exactly what was proposed by SCI (differential
signalling with true current steering) to avoid those
signal integrity blows like 3-rd RIMM, and save years
in time-to-market.
In short, one should not take the marketing terminology
as a gospel and assume ideal conditions in lame attempts
to justify unjustifiable. It is apparent
that the "Rambus signaling technology" is a collection
of heavy compromises to achieve a barely working signal
quality.
- Ali |
