The number of times to erase flash reaches a million times by FS or NVMC

Question

hi, 
 I am testing the flash erasing times of nRF52832, and i find that the minimum times of flash erase\write are 10000 documented in the data sheet. 
 I konw the times are tested at very high temperature(non-normal temperature), so I 'm curious about how many times of flash erase/write are at room tempetature. 
 The FS and NVMC are programing and downloaded in the nRF52832 chip. 
 The erase times can get by J-link RTT Viewer, and the codes generate random numbers and write them to flash. 
 Just pay attention to this times of flash erase (the red arrow is marked) . 
 
 Regarding the question： 
 1) are the times right? Why is it that the flash is still not damaged after so many erases? 
 2) if not, could i know the times of flash erase/write at room tempetature? 
 3) are the codes right attached the end of the page? 
 
 FS code log: run many times and reset by manual (erase times more than a million times in total)

NVMC code log: erase times more than a million times in total

FS code: 
 while(true)
	{
 NRF_LOG_FLUSH();
 NRF_LOG_INFO("
 =============================================== ");
 
 uint8_t p_buff[RANDOM_BUFF_SIZE];
 uint8_t length = random_vector_generate(p_buff,RANDOM_BUFF_SIZE);
 
 NRF_LOG_INFO("Write Random Vector:");
 NRF_LOG_HEXDUMP_INFO(p_buff, length);
 NRF_LOG_INFO("");
 NRF_LOG_FLUSH();
 
 nrf_delay_ms(20);
 
 memcpy(my_fs_info.write_data, p_buff, length);
 my_fs_info.write_len = length;
 nrf_delay_ms(100);
 
 
 // flash write
 if((my_fs_info.need_write == true) && (my_fs_info.busy == false)&&(reverse_flag == false))
 { 
 if((write_err_cnt > 100) || (erase_err_cnt > 100))
 {
 NRF_LOG_INFO(" ==== write enough!!!! ====");
 NRF_LOG_INFO(" ==== write enough!!!! ====");
 continue;
 }
 if(my_fs_info.write_len == 0)
 {
 continue;
 }
 my_fs_info.busy = true; 
 my_fs_info.need_write = false;
 my_fs_info.need_read = true;
 reverse_flag = true;
 
 NRF_LOG_INFO("Address: %08X", FLASH_START_ADDR);
 rc = nrf_fstorage_write(&my_fs, FLASH_START_ADDR, &my_fs_info.write_data, my_fs_info.write_len, NULL);
 if(rc != NRF_SUCCESS)
 {
 write_err_cnt += 1;
 NRF_LOG_INFO("== fs write error: 0x%x",rc);	
 }
 else
 {
 write_cnt += 1;
 NRF_LOG_INFO(" === write count: %d", write_cnt);
 }
 }
 
 // flash read
 if((my_fs_info.need_read == true)&&(my_fs_info.busy == false))
 {
 my_fs_info.need_read = false;
 
 rc = nrf_fstorage_read(&my_fs, FLASH_START_ADDR, &my_fs_info.read_data,256);
 
 if(rc != NRF_SUCCESS)
 {
 NRF_LOG_INFO("0x%x",rc);
 }
 else
 {
 NRF_LOG_INFO("Read Flash Data:");
 NRF_LOG_HEXDUMP_INFO(my_fs_info.read_data, RANDOM_BUFF_SIZE); 
 }
 }
 
 // flash erase
 if((my_fs_info.need_write == false) && (my_fs_info.busy == false)&&(reverse_flag == true))
 { 
 if((write_err_cnt > 100) || (erase_err_cnt > 100))
 {
 NRF_LOG_INFO(" ==== erase enough!!!! ====");
 NRF_LOG_INFO(" ==== erase enough!!!! ====");
 continue;
 }
 my_fs_info.need_read = true;
 my_fs_info.need_write = true;
 my_fs_info.busy = true;
 reverse_flag = false;
 rc = nrf_fstorage_erase(&my_fs, FLASH_START_ADDR, 1, NULL);
 if(rc != NRF_SUCCESS)/
 {
 erase_err_cnt += 1;
 NRF_LOG_INFO("== fs erase error: 0x%x",rc);
 }
 else
 {
 erase_cnt += 1;
 NRF_LOG_INFO(" === erase count: %d", erase_cnt);
 }
 }

 idle_state_handle();
	}
}

NVMC code: 
 while(status) 
{ 
 addr = 0x00072000;
 nrf_nvmc_page_erase(addr);
 nrf_nvmc_write_word(addr,src_value);
 pdat = (uint32_t *)addr;
 
 NRF_LOG_INFO("0x%x was read from flash
", *pdat);
 
 if((*pdat) != src_value)
 {
 status = false;
 
 NRF_LOG_INFO(" ===== Flash Error !!! =====");
 NRF_LOG_FLUSH();
 break;
 }
 opt_cnt += 1;
 NRF_LOG_INFO(" ===== flash operation : %d", opt_cnt); 
 NRF_LOG_FLUSH();
 }
} 
 
 best wishes, 
 thanks.

ovrebekk · Answer

Hi 
 1) are the times right? Why is it that the flash is still not damaged after so many erases? 
 The quoted numbers in the datasheet are the worst case values, which has to take things like different temperatures, different supply voltages, production variations and the effects of aging into account. Tests on a small number of devices in lab conditions will always show better results. 
 And even if no errors can be detected now you might still experience reliability issues after some time has passed, or if you change the operating environment. 
 2) if not, could i know the times of flash erase/write at room tempetature? 
 We don't test explicitly for this, so we don't have any values to share unfortunately. The general recommendation is to design for the worst case (minimum) listed in the documentation, and make sure your application will work well even if the worst case hits. 
 3) are the codes right attached the end of the page? 
 They code looks OK to me. Did you measure the timing of the erase call? 
 It could take as much as 20-80ms to run a single page erase, and as such running more than a million page erase cycles should be quite time consuming to test. 
 Is there any particular reason you need such a high amount of erase cycles? Are you continuously storing large amounts of data to flash? 
 Best regards Torbjørn

The number of times to erase flash reaches a million times by FS or NVMC

Top Replies