Skip to content

Commit e8bbbaf

Browse files
author
Filip Jagodzinski
committed
Test: LP_Ticker: Increase the UART flush wait
1 parent fa7eb7d commit e8bbbaf

File tree

1 file changed

+11
-12
lines changed

1 file changed

+11
-12
lines changed

TESTS/mbed_hal/lp_ticker/main.cpp

Lines changed: 11 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -47,21 +47,20 @@ ticker_irq_handler_type prev_handler;
4747

4848
#define LP_TICKER_OV_LIMIT 4000
4949

50-
/* Flush serial buffer before deep sleep
50+
/* To prevent a loss of Greentea data, the serial buffers have to be flushed
51+
* before the UART peripheral shutdown. The UART shutdown happens when the
52+
* device is entering the deepsleep mode or performing a reset.
5153
*
52-
* Since deepsleep() may shut down the UART peripheral, we wait for some time
53-
* to allow for hardware serial buffers to completely flush.
54+
* With the current API, it is not possible to check if the hardware buffers
55+
* are empty. However, it is possible to determine the time required for the
56+
* buffers to flush.
5457
*
55-
* Take NUMAKER_PFM_NUC472 as an example:
56-
* Its UART peripheral has 16-byte Tx FIFO. With baud rate set to 9600, flush
57-
* Tx FIFO would take: 16 * 8 * 1000 / 9600 = 13.3 (ms). So set wait time to
58-
* 20ms here for safe.
59-
*
60-
* This should be replaced with a better function that checks if the
61-
* hardware buffers are empty. However, such an API does not exist now,
62-
* so we'll use the busy_wait_ms() function for now.
58+
* Assuming the biggest Tx FIFO of 128 bytes (as for CY8CPROTO_062_4343W)
59+
* and a default UART config (9600, 8N1), flushing the Tx FIFO wold take:
60+
* (1 start_bit + 8 data_bits + 1 stop_bit) * 128 * 1000 / 9600 = 133.3 ms.
61+
* To be on the safe side, set the wait time to 150 ms.
6362
*/
64-
#define SERIAL_FLUSH_TIME_MS 20
63+
#define SERIAL_FLUSH_TIME_MS 150
6564

6665
void busy_wait_ms(int ms)
6766
{

0 commit comments

Comments
 (0)