Transcript Integrated Placement and Skew Optimization for Rotary Clocking

Integrated Placement and Skew
Optimization for Rotary Clocking
A paper by: Ganesh Venkataraman, Student Member, IEEE,
Jiang Hu,
Member, IEEE, and Frank (Ying) Liu, Member,
IEEE
Presented By: Eric Tramel
Global Distribution Scheme



Each rotary section consists of
a pair of cross connected
differential transmission line
circles which can be linked
together to cover the entirety
of the die.
This travelling wave clock
approach allows us to
physically predict at what point
on the transmission line the
clock will be at a particular
phase
Closed loop line helps save on
power for a global clock grid.
All we need is a single or
sporadic clock injection into
the system and the clock will
recirculate throughout the
system.
Advantages of a Rotary Clock
Scheme...

Low Power Dissipation: Because of the lack of a single
clock source driving the entire chip, we can avoid the power
dissipation caused by the charging/discharging over the chip's
capacitive load. This is achieved by the closed loop geometry as well
as inverters placed between phase-opposite locations within individual
rings (also helps reduce skew variation). These features have been
shown to reduce power consumption of the clock network by up to
70%.
 Low Skew Variation: The consistent geometry of the
rotary clock mesh allows us to accurately predict the phase of
the clock signal on the transmission lines at any point. Test chip
evaluation has shown skew variation as low as 5.5ps at a clock
speed of 950 Mhz (0.52%)!
...and then Downsides.

Not efficient for zero skew designs:
Because the zero skew points occur at exactly one position on
each ring, it would be entirely inefficient to use only this single
portion of the ring. In order to use the rotary clock scheme to its
full potential, calculations must be made in order to tap the clock
ring at different phase locations and buffers added in order to
ensure proper edge arrival.

Requires a more complicated CAD flow:
“Unlike the intentional skew design in the conventional clocking
technology where no restrictions are imposed on the flip-flop
locations, the skew at each flip-flop has to be matched with a
specific location at the rotary clock ring. This requirement forms
a difficult chicken-and-egg problem: the flip-flop placement
depends on skew optimization while it is well known that skew
optimization depends on flip-flop locations.”
Methodology for Overcoming the
Design Hurdles