......................DOC #: SP-AP-0052 (Rev. AA) Page 15 of 31
CL ................................................... Crystal load capacitance
CLe .........................................Actual loading seen by crystal
using standard value trim capacitors
Ce .....................................................External trim capacitors
Cs ............................................. Stray capacitance (terraced)
Ci .......................................................... Internal capacitance
(lead frame, bond wires, etc.)
Dial-A-Frequency
®
(CPU and PCIEX)
This feature allows the user to over-clock their system by
slowly stepping up the CPU or SRC frequency. When the
programmable output frequency feature is enabled, the CPU
and SRC frequencies are determined by the following
equation:
Fcpu = G * N/M or Fcpu=G2 * N, where G2 = G / M.
• “N” and “M” are the values programmed in Programmable
Frequency Select N-Value Register and M-Value Register,
respectively.
• “G” stands for the PLL Gear Constant, which is determined
by the programmed value of FS[E:A]. See Table ,
Frequency Select Table for the Gear Constant for each
Frequency selection. The PCI Express only allows user
control of the N register, the M value is fixed and
documented in Table , Frequency Select Table.
In this mode, the user writes the desired N and M values into
the DAF I2C registers. The user cannot change only the M
value and must change both the M and the N values at the
same time, if they require a change to the M value. The user
may change only the N value.
Associated Register Bits
• CPU_DAF Enable – This bit enables CPU DAF mode. By
default, it is not set. When set, the operating frequency is
determined by the values entered into the CPU_DAF_N
register. Note that the CPU_DAF_N and M register must
contain valid values before CPU_DAF is set. Default = 0,
(No DAF).
• CPU_DAF_N – There are nine bits (for 512 values) to
linearly change the CPU frequency (limited by VCO range).
Default = 0, (0000). The allowable values for N are detailed
in Table , Frequency Select Table.
• CPU DAF M – There are 7 bits (for 128 values) to linearly
change the CPU frequency (limited by VCO range). Default
= 0, the allowable values for M are detailed in Table ,
Frequency Select Table
• SRC_DAF Enable – This bit enables SRC DAF mode. By
default, it is not set. When set, the operating frequency is
determined by the values entered into the SRC_DAF_N
register. Note that the SRC_DAF_N register must contain
valid values before SRC_DAF is set. Default = 0, (No DAF).
• SRC_DAF_N – There are nine bits (for 512 values) to
linearly change the CPU frequency (limited by VCO range).
Default = 0, (0000). The allowable values for N are detailed
in Table , Frequency Select Table.
Smooth Switching
The device contains one smooth switch circuit that is shared
by the CPU PLL and SRC PLL. The smooth switch circuit
ensures that when the output frequency changes by
overclocking, the transition from the old frequency to the new
frequency is a slow, smooth transition containing no glitches.
The rate of change of output frequency when using the smooth
switch circuit is less than 1 MHz/0.667 s. The frequency
overshoot and undershoot is less than 2%.
The Smooth Switch circuit assigns auto or manual. In Auto
mode, clock generator assigns smooth switch automatically
when the PLL does overclocking. For manual mode, assign
the smooth switch circuit to PLL via Smbus. By default the
smooth switch circuit is set to auto mode. PLL can be
over-clocked when it does not have control of the smooth
switch circuit but it is not guaranteed to transition to the new
frequency without large frequency glitches.
Do not enable over-clocking and change the N values of both
PLLs in the same SMBUS block write and use smooth switch
mechanism on spread spectrum on/off.
PD_RESTORE
If a ‘0’ is set for Byte 0 bit 0 then, upon assertion of PWRDWN#
LOW, the SL28504 initiates a full reset. The result of this is that
the clock chip emulates a cold power on start and goes to the
“Latches Open” state. If the PD_RESTORE bit is set to a ‘1’
then the configuration is stored upon PWRDWN# asserted
LOW. Note that if the iAMT bit, Byte 0 bit 3, is set to a ‘1’ then
the PD_RESTORE bit must be ignored. In other words, in Intel
iAMT mode, PWRDWN# reset is not allowed.
PWRDWN# (Power down) Clarification
The CKPWRGD/PWRDWN# pin is a dual-function pin. During
initial power up, the pin functions as CKPWRGD. Once
CKPWRGD has been sampled HIGH by the clock chip, the pin
assumes PD# functionality. The PD# pin is an asynchronous
active LOW input used to shut off all clocks cleanly before
shutting off power to the device. This signal is synchronized
internally to the device before powering down the clock
synthesizer. PD# is also an asynchronous input for powering
up the system. When PD# is asserted LOW, clocks are driven
to a LOW value and held before turning off the VCOs and the
crystal oscillator.
PWRDWN# (Power down) Assertion
When PD is sampled HIGH by two consecutive rising edges
of CPUC, all single-ended outputs will be held LOW on their
next HIGH-to-LOW transition and differential clocks must held
LOW. When PD mode is desired as the initial power on state,
PD must be asserted HIGH in less than 10 s after asserting
CKPWRGD.
PWRDWN# Deassertion
The power up latency is less than 1.8 ms. This is the time from
the deassertion of the PD# pin or the ramping of the power
supply until the time that stable clocks are generated from the
clock chip. All differential outputs stopped in a three-state
condition, resulting from power down are driven high in less
than 300 s of PD# deassertion to a voltage greater than
200 mV. After the clock chip’s internal PLL is powered up and
locked, all outputs are enabled within a few clock cycles of
