Improve sf32lb52 low power mode stability.#364
Conversation
cameled
force-pushed
the
deepsleep
branch
2 times, most recently
from
e2fede0 to
f7fc0ea
Compare
cameled
changed the title
cameled
requested review from
Jiyun-Yang,
gangheivt,
gmarull and
rabbitsaviola
| // Reset HR pin, clear IE, PE | ||
| hwp_pinmux1->PAD_PA09 &= ~((1 << 6) | (1 << 4)); | ||
| hwp_pinmux1->PAD_PA20 &= ~((1 << 6) | (1 << 4)); | ||
| hwp_pinmux1->PAD_PA25 &= ~((1 << 6) | (1 << 4)); |
There was a problem hiding this comment.
Leak is from PA09 or PA20(forget which one). As I tested, after reset that pin, current drop 150uA.
|
Measured average current when connected to a phone (blue led is desoldered): 
I observe display turning randomly to black, so something's broken there, like if it is being disabled or something, can you check? (using BB2) |
|
Original performance data is measured without hxt48 change. Looks hxt48 enable is a new point we get, but not the final. |
Could you also check display problems? |
cameled
force-pushed
the
deepsleep
branch
2 times, most recently
from
940e724 to
6d53569
Compare
black or dim? I see it will enter dim state. |
|
please rebase |
The AON IRQ pending issue have been fixed, also add a ipc queue idle check to avoid LCPU enter incorrect state. The unstable wakeup time issue still remain, sometimes need nearly 8ms to wakeup. Thanks for xiaoming's support. Signed-off-by: HaiLong Yang <cameled@outlook.com>
Use enabled flag to avoid stop mode counter inconsistent. Signed-off-by: HaiLong Yang <cameled@outlook.com>
Signed-off-by: HaiLong Yang <cameled@outlook.com>
Set RC32K as LP clock, which will reduce the deepsleep wakeup time. Signed-off-by: HaiLong Yang <cameled@outlook.com>
|
@gmarull The black issue fixed. |
I can see firmware is stable now, average current is: 
|
2 participants
This improve the sf32lb52 low power mode stability.
Low power current on big board is mainly from the blue led, nearly 2 mA. Also find current leak on HR pin, about 150uA. Current low power performance as below. (still have some IO leak?)
In deepsleep mode, HXT48 clock should be enable. its effective the LCPU/BLE and wakeup time. After HXT48 clock enable, BLE worked (not full tested), and wakeup time drop to 1ms.
Remain work: