Higher current draw in System On low-power mode after ESB PTX than after ESB PRX

[NewtoM]

Dear All,

while evaluating low-power options in the context of wireless sensors, I've stumbled upon a strange behavior with the nRF52832 using ESB from SDK 14.2. Here's the outline of my test code:

#include "nrf_delay.h"
#include "nrf_drv_clock.h"
#include "nrf_drv_power.h"
#include "nrf_esb.h"
#include "nrf_pwr_mgmt.h"
#include "nrf.h"

static nrf_esb_config_t const m_esbConfig_TX=
{
  .protocol           = NRF_ESB_PROTOCOL_ESB_DPL,
  .mode               = NRF_ESB_MODE_PTX,
  .event_handler      = EsbEventHandler,
  .bitrate            = NRF_ESB_BITRATE_2MBPS,
  .crc                = NRF_ESB_CRC_16BIT,
  .tx_output_power    = NRF_ESB_TX_POWER_0DBM,
  .retransmit_delay   = 250,
  .retransmit_count   = 100,
  .tx_mode            = NRF_ESB_TXMODE_MANUAL,
  .radio_irq_priority = 1,
  .event_irq_priority = 2,
  .payload_length     = NRF_ESB_MAX_PAYLOAD_LENGTH,
  .selective_auto_ack = false
};

static nrf_esb_config_t const m_esbConfig_RX=
{
  .protocol           = NRF_ESB_PROTOCOL_ESB_DPL,
  .mode               = NRF_ESB_MODE_PRX,
  .event_handler      = EsbEventHandler,
  .bitrate            = NRF_ESB_BITRATE_2MBPS,
  .crc                = NRF_ESB_CRC_16BIT,
  .tx_output_power    = NRF_ESB_TX_POWER_0DBM,
  .retransmit_delay   = 250,
  .retransmit_count   = 3,
  .tx_mode            = NRF_ESB_TXMODE_AUTO,
  .radio_irq_priority = 1,
  .event_irq_priority = 2,
  .payload_length     = NRF_ESB_MAX_PAYLOAD_LENGTH,
  .selective_auto_ack = false
};


int main(void)
{
  ret_code_t retCode= NRF_SUCCESS;

  nrf_pwr_mgmt_init();

  nrf_drv_clock_init();
  nrf_drv_clock_lfclk_request(NULL);
  nrf_drv_clock_hfclk_request(NULL);
  
  nrf_esb_set_prefixes(m_esbAddressPrefix, 1);
  nrf_esb_set_base_address_0(m_esbAddressBase0);
  nrf_esb_set_base_address_1(m_esbAddressBase1);
  nrf_esb_set_rf_channel(10);
  
  // initialize radio in the specified ESB mode (use m_esbConfig_RX or m_esbConfig_TX)
  nrf_esb_config_t esbConfig= m_esbConfig_TX;
  nrf_esb_init(&esbConfig);
  
  if (esbConfig.mode == NRF_ESB_MODE_PTX)
  {
    nrf_esb_payload_t payloadTx= {0};
    payloadTx.length= 16;
    nrf_esb_write_payload(&payloadTx);

    nrf_delay_ms(1000);
    nrf_esb_start_tx();
    nrf_delay_ms(1000);
  }
  
  else if (esbConfig.mode == NRF_ESB_MODE_PRX)
  {
    nrf_delay_ms(1000);
    nrf_esb_start_rx();
    nrf_delay_ms(1000);
    
    nrf_esb_stop_rx();
  }

  nrf_esb_disable();

  nrf_drv_clock_hfclk_release();

  while (1)
  {
    nrf_pwr_mgmt_run();
  }
}

The last few lines put the system into System On low-power mode, and that's where it becomes interesting. It seems if I use the radio in RX mode and then disable it, the system draws about 1.4 µA (3V) current, whereas if I use the radio in TX mode and then disable it, the current drawn is about 8 µA. I briefly looked into the ESB implementation, but I can't find the bug (if any). Here two graphs showing the problem:

Things I tried:

• disabling all interrupts manually right after the nrf_esb_disable() call --> no difference;
• manually releasing PPI channels that may have been left configired by ESB --> no difference;
• setting m_address_hang_fix_enable in nrf_esb.c to false --> no difference;
• after using the ESB in PTX mode, re-setting it to PRX, then disable --> no difference.

I would really like to sort out this problem. Any help would be appreciated!

Thanks,

Tamas

[NewtoM]

The issue is still relevant. While 1.4 µA is lower than the "sleep" current of competing devices, 8 µA would rule out the Nordic chip in our intended scenario.

[NewtoM]

In the meantime I repeated the test with the nRF52840 Preview DK (External 3V supply, nRF only setting). The problem DOES exist there, too. Current consumption after disabling ESB and calling nrf_pwr_mgmt_run() in a loop:

• 17 µA if the ESB was previously initialized as PTX, and
• 2.7 µA if the ESB was previously only initalized as PRX.

In System Off mode the current consumption drops to 1.8 µA in both cases.

[ovrebekk]

Hi Tamas

It sounds to me like you have hit product anomaly #78, which makes the timer module draw excessive current in sleep unless the SHUTDOWN task is triggered. 

Can you try to shut down the timer before entering sleep?

Like this:

NRF_ESB_SYS_TIMER->TASKS_SHUTDOWN = 1;

Best regards
Torbjørn Øvrebekk

[NewtoM]

Dear Torbjørn,

you're right. I inserted the proposed line right after nrf_esb_disable(); and my current draw dropped to 2.7 µA in "sleep" mode after using ESB in TX mode.

Thank you very much!

Tamas

[Stian Røed Hafskjold]

The bugfix timer that was introduced in SDK14 also needs to be shut down before sleep in order to get lower consumption.

NRF_ESB_BUGFIX_TIMER->TASKS_SHUTDOWN = 1;