Two functional clock inputs are
defined for the ESC: CLK25 (25MHz) and CLK100 (100MHz).
EtherCAT functional clock inputs CLK25
and CLK100 are sourced by the MCU clocking module either using SYSPLL or AUXPLL. At
the SoC level, you have multiple options to choose from regarding what inputs are
used for the SYSPLL or AUXPLL.
- Due to the 25ppm requirement for
EtherCAT, an external oscillator of 25MHz is suggested to be used as the main
clock source.
- A less accurate clock than 25ppm
limits the ability of the ESC to act as the network reference clock. For
practical reasons, clock accuracy must be the same or better than the Ethernet
clock source that is 50ppm.
Note:
- CLK25 and CLK100 are used for
the ESC core operation. These two clocks and the clocks to the PHY must
have a common source, which establishes a deterministic phase
relationship between PHY clocks and ESC clocks.
- In MII mode, 25MHz RX clock
is sourced by the PHY while the TX clock from PHY is optional (which
effectively saves a pin). The phase differential between the PHY TX clock
and CLK25 can be compensated to the TX data and TX EN using the manual
compensation mode in increments of 10ns.
Figure 31-10 shows the clocks connectivity. Refer to the System Control and Interrupts
chapter for additional details of the PLL source selections and clock dividers.
In the case
of a low-accuracy clock (up to 100ppm) being used, a clock accuracy of 25ppm cannot
be feasible, then the following restrictions apply:
- RX FIFO size cannot be reduced
lower than 7 (default for 100ppm), which in turn affects the latency of the
transfer as the transfer starts after half of the FIFO full threshold, thus
resulting in a RX Delay.
- This ECS cannot be used as the
first slave from the
master in the network,
as typically the first slave from the master is treated as
the reference clock for the distributed clocks functionality.
- The number of iterations required
for synchronizing clocks (specially the drift computation) increases.