drivers: clock_control: nrf2: improve audiopll clock calculations #97735
Conversation
zephyrbot
requested review from
anangl,
bjarki-andreasen,
jaz1-nordic,
kl-cruz,
lstnl,
magp-nordic,
masz-nordic,
nika-nordic and
nordic-krch
Improve the Audio PLL clock frequency calculations to provide a more accurate clock frequency. Signed-off-by: Victor Brzeski <[email protected]>
vbrzeski
force-pushed
the
audiopll-freq
branch
from
563f97a to
768819d
Compare
|
|
I'm actually unsure if this formula in the source code is correct. The standard div value overflows for 48k |
| case 8: | ||
| return AUDIOPLL_DIV_8; | ||
| case 12: | ||
| return AUDIOPLL_DIV_12; |
There was a problem hiding this comment.
Prescaler div 12 was the only used and valid prescaler when this driver was written, has that changed?
@tmon-nordic any input? |
Where do you see overflow? Code uses
It is important to note that while the comment |
| switch (divider) { | ||
| case 0: | ||
| return AUDIOPLL_DIV_DISABLED; | ||
| case 6: | ||
| return AUDIOPLL_DIV_6; | ||
| case 8: | ||
| return AUDIOPLL_DIV_8; | ||
| case 12: | ||
| return AUDIOPLL_DIV_12; | ||
| case 16: | ||
| return AUDIOPLL_DIV_16; | ||
| default: | ||
| __ASSERT(divider <= 4, "invalid audiopll divider"); | ||
| return (enum audiopll_prescaler_div)divider; | ||
| } |
There was a problem hiding this comment.
The values seem to originate from nrfs header, but I have no idea why nrfs makes them available. I would prefer if AudioPLL just continues to use hardcoded AUDIOPLL_DIV_12 which is known to work.
|
With corrected comparison script, the results are as follows:
Therefore new formula is slightly better on accuracy. |
7 participants
Improve the Audio PLL clock frequency calculations to provide a more accurate clock frequency.
These formulas were derived with a Wolfram Alpha query using the long-form formula in the source comments.