lpcware

Picking up the best pll configuration automatically

Discussion created by lpcware Employee on Jun 15, 2016
Latest reply on Jun 15, 2016 by lpcware
Content originally posted in LPCWare by farcaller on Fri Jun 13 07:51:43 MST 2014
I'm trying to write an algorithm that would pick the best available PLL0 configuration for the user, given F_in and F_cclk. For now, I don't care about USB clocking.

Let's say the user has mbed board (external 12MHz crystal) that he wants clocked at 100MHz, so the input is:

F_in = 12MHz
F_cclk = 100MHz

The manual says that I should pick higher F_ref if possible for a more stable PLL, for low frequency input clock. What is considered a "low frequency", though?

Also, I can get quite flexible with F_cco value, given that it's later passed through a divisor. So I have two flexibility points here:

[list]
  [*] prefer higher M/N pair to lower M/N (i.e. prefer higher F_ref to lower F_ref)
  [*] prefer higher F_cco/divisor to lower F_cco/divisor
[/list]

Is there any generic rule which one of those I should prefer? What is better, m=50, n=3, divisor=4, or, say, m=325, n=26, divisor=3?

Outcomes