All clock sources in the system can be run in an on-demand mode: the clock source is in a stopped state unless a peripheral is requesting the clock source. Clock requests propagate from the peripheral, through the GCLK, to the clock source. If one or more peripheral is using a clock source, the clock source will be started/kept running. As soon as the clock source is no longer needed and no peripheral has an active request, the clock source will be stopped until requested again.
The clock request can reach the clock source only if the peripheral, the generic clock and the clock from the Generic Clock Generator in-between are enabled. The time taken from a clock request being asserted to the clock source being ready is dependent on the clock source startup time, clock source frequency as well as the divider used in the Generic Clock Generator. The total startup time Tstart from a clock request until the clock is available for the peripheral is between:
Tstart_max = Clock source startup time + 2 × clock source periods + 2 × divided clock source periods
Tstart_min = Clock source startup time + 1 × clock source period + 1 × divided clock source period
The time between the last active clock request stopped and the clock is shut down, Tstop, is between:
Tstop_min = 1 × divided clock source period + 1 × clock source period
Tstop_max = 2 × divided clock source periods + 2 × clock source periods
The On-Demand function can be disabled individually for each clock source by clearing the ONDEMAND bit located in each clock source controller. Consequently, the clock will always run whatever the clock request status is. This has the effect of removing the clock source startup time at the cost of power consumption.
The clock request mechanism can be configured to work in Standby mode by setting the RUNSDTBY bits of the modules as shown in the figure above.