Re: [linux-sunxi] [PATCH] clk: sunxi-ng: sun8i: r40: Use sigma delta modulation for audio PLL

From: Chen-Yu Tsai
Date: Tue Aug 25 2020 - 10:46:51 EST


On Tue, Aug 25, 2020 at 9:11 PM Jernej Skrabec <jernej.skrabec@xxxxxxxx> wrote:
>
> Audio cores need specific clock rates which can't be simply obtained by
> adjusting integer multipliers and dividers. HW for such cases supports
> delta-sigma modulation which enables fractional multipliers.
>
> Port H3 delta-sigma table to R40. They have identical audio PLLs.
>
> Signed-off-by: Jernej Skrabec <jernej.skrabec@xxxxxxxx>
> ---
> drivers/clk/sunxi-ng/ccu-sun8i-r40.c | 37 ++++++++++++++++++----------
> 1 file changed, 24 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/clk/sunxi-ng/ccu-sun8i-r40.c b/drivers/clk/sunxi-ng/ccu-sun8i-r40.c
> index 23bfe1d12f21..84153418453f 100644
> --- a/drivers/clk/sunxi-ng/ccu-sun8i-r40.c
> +++ b/drivers/clk/sunxi-ng/ccu-sun8i-r40.c
> @@ -45,18 +45,29 @@ static struct ccu_nkmp pll_cpu_clk = {
> * the base (2x, 4x and 8x), and one variable divider (the one true
> * pll audio).
> *
> - * We don't have any need for the variable divider for now, so we just
> - * hardcode it to match with the clock names
> + * With sigma-delta modulation for fractional-N on the audio PLL,
> + * we have to use specific dividers. This means the variable divider
> + * can no longer be used, as the audio codec requests the exact clock
> + * rates we support through this mechanism. So we now hard code the
> + * variable divider to 1. This means the clock rates will no longer
> + * match the clock names.
> */
> #define SUN8I_R40_PLL_AUDIO_REG 0x008
>
> -static SUNXI_CCU_NM_WITH_GATE_LOCK(pll_audio_base_clk, "pll-audio-base",
> - "osc24M", 0x008,
> - 8, 7, /* N */
> - 0, 5, /* M */
> - BIT(31), /* gate */
> - BIT(28), /* lock */
> - CLK_SET_RATE_UNGATE);
> +static struct ccu_sdm_setting pll_audio_sdm_table[] = {
> + { .rate = 22579200, .pattern = 0xc0010d84, .m = 8, .n = 7 },
> + { .rate = 24576000, .pattern = 0xc000ac02, .m = 14, .n = 14 },

The user manual has an additional requirement: 3 <= N/M <= 21.
Though it then says 72 <= 24*N/P <= 504. Not sure which one is
right...

Did you run into any glitches or audio distortions?

ChenYu