Rare counter error

[CheMax]

Hello everybody,

I ran into an unpleasant moment: sometimes the timer counts incorrectly in the counter mode.

What is used:

• Custom board with BC832 (Fanstel, NRF52832xxAA);
• SDK v15.0.0. SD 6.1.0;
• Segger Embedded Studio;

Additional components:

• ADS1292 (SPI);
• two I2C devices (charger and accelerometr);

What should be done:

• Read data from ADS (1 kHz) and read data from i2c devices (~100 Hz);
• Generate a square wave at a frequency of 25 Hz based on the data ready signal from the ADS1292;

The following settings have been used for the operation of the SPI:

void AdsInitHW (void)
{
  // Настройка входа готовности данных
  NRF_GPIO->PIN_CNF[AFE_DRDYn] = 0x00;
  NRF_GPIO->PIN_CNF[AFE_DRDYn] = (GPIO_PIN_CNF_DIR_Input << GPIO_PIN_CNF_DIR_Pos) | (GPIO_PIN_CNF_PULL_Disabled << GPIO_PIN_CNF_PULL_Pos);
  
  // ADS Dready
  NRF_GPIOTE->CONFIG[AFE_DRDY_GPIOTE] |= GPIOTE_CONFIG_MODE_Event << GPIOTE_CONFIG_MODE_Pos;
  NRF_GPIOTE->CONFIG[AFE_DRDY_GPIOTE] |= AFE_DRDYn<<GPIOTE_CONFIG_PSEL_Pos;
  NRF_GPIOTE->CONFIG[AFE_DRDY_GPIOTE] |= GPIOTE_CONFIG_POLARITY_HiToLo << GPIOTE_CONFIG_POLARITY_Pos;

  // (Chip Select)  
  NRF_GPIOTE->CONFIG[AFE_CS_GPIOTE] |= GPIOTE_CONFIG_MODE_Task << GPIOTE_CONFIG_MODE_Pos;
  NRF_GPIOTE->CONFIG[AFE_CS_GPIOTE] |= AFE_CSn<<GPIOTE_CONFIG_PSEL_Pos;
  NRF_GPIOTE->TASKS_SET[AFE_CS_GPIOTE] = 1; // ChipSelect not active

  NRF_PPI->CH[AFE_PPI_Ch2].EEP	= (uint32_t) &AFE_SPI->EVENTS_ENDTX;
  NRF_PPI->CH[AFE_PPI_Ch2].TEP	= (uint32_t) &NRF_GPIOTE->TASKS_SET[AFE_CS_GPIOTE];
  
  
  AFE_SPI->ENABLE	= SPIM_ENABLE_ENABLE_Enabled;
  
  AFE_SPI->PSEL.SCK   = ((AFE_SCLK<<SPIM_PSEL_SCK_PIN_Pos)	| (SPIM_PSEL_SCK_CONNECT_Connected  << SPIM_PSEL_SCK_CONNECT_Pos));
  AFE_SPI->PSEL.MISO	= ((AFE_DOUT<<SPIM_PSEL_MISO_PIN_Pos)	| (SPIM_PSEL_MISO_CONNECT_Connected << SPIM_PSEL_MISO_CONNECT_Pos));
  AFE_SPI->PSEL.MOSI	= ((AFE_DIN<<SPIM_PSEL_MOSI_PIN_Pos)	| (SPIM_PSEL_MOSI_CONNECT_Connected << SPIM_PSEL_MOSI_CONNECT_Pos));
  AFE_SPI->CONFIG     = SPIM_CONFIG_CPHA_Trailing<<SPIM_CONFIG_CPHA_Pos;
  AFE_SPI->FREQUENCY	= SPIM_FREQUENCY_FREQUENCY_K250;

  AFE_SPI->INTENSET = SPIM_INTENSET_STOPPED_Set << SPIM_INTENSET_STOPPED_Pos;
  AFE_SPI->INTENSET = SPIM_INTENSET_ENDRX_Set   << SPIM_INTENSET_ENDRX_Pos;
  AFE_SPI->INTENSET = SPIM_INTENSET_END_Set     << SPIM_INTENSET_END_Pos;
  AFE_SPI->INTENSET = SPIM_INTENSET_ENDTX_Set   << SPIM_INTENSET_ENDTX_Pos;
  AFE_SPI->INTENSET = SPIM_INTENSET_STARTED_Set << SPIM_INTENSET_STARTED_Pos;

  NVIC_EnableIRQ(AFE_IRQ);
}

void AdsSetFrameConfig (uint8_t sample_size, uint8_t frame_size, uint8_t frame_sync, uint8_t *data_ptr)
{
  
  NRF_PPI->CH[AFE_PPI_Ch1].EEP	= (uint32_t) &NRF_GPIOTE->EVENTS_IN[AFE_DRDY_GPIOTE];
  NRF_PPI->CH[AFE_PPI_Ch1].TEP	= (uint32_t) &NRF_GPIOTE->TASKS_CLR[AFE_CS_GPIOTE];
  NRF_PPI->FORK[AFE_PPI_Ch1].TEP= (uint32_t) &AFE_SPI->TASKS_START;
  
  NRF_PPI->CH[AFE_PPI_Ch2].EEP	= (uint32_t) &AFE_SPI->EVENTS_ENDRX;
  NRF_PPI->FORK[AFE_PPI_Ch2].TEP= (uint32_t) &AFE_EV_CNT->TASKS_COUNT;
  
  NRF_PPI->CHENSET = 1 << AFE_PPI_Ch1;
  NRF_PPI->CHENSET = 1 << AFE_PPI_Ch2;

  AFE_SPI->RXD.LIST   = SPIM_RXD_LIST_LIST_ArrayList;
  AFE_SPI->RXD.PTR    = (uint32_t)data_ptr;
  AFE_SPI->RXD.MAXCNT = sample_size;

  AFE_SPI->TXD.PTR    = NULL;
  AFE_SPI->TXD.MAXCNT = 0;
  
  AFE_EV_CNT->MODE    = TIMER_MODE_MODE_LowPowerCounter<<TIMER_MODE_MODE_Pos;
  AFE_EV_CNT->BITMODE = TIMER_BITMODE_BITMODE_08Bit;                         
  AFE_EV_CNT->CC[0]   = frame_size; // 1 frame = 100 samples                                          
  AFE_EV_CNT->CC[1]   = frame_sync; // for sync run another stream
  AFE_EV_CNT->SHORTS  = TIMER_SHORTS_COMPARE0_CLEAR_Enabled << TIMER_SHORTS_COMPARE0_CLEAR_Pos;
  
  AFE_EV_CNT->INTENSET = TIMER_INTENSET_COMPARE0_Enabled << TIMER_INTENSET_COMPARE0_Pos;
  AFE_EV_CNT->INTENSET = TIMER_INTENSET_COMPARE1_Enabled << TIMER_INTENSET_COMPARE0_Pos;
  NVIC_SetPriority(AFE_EV_CNT_IRQ, AFE_EV_CNT_IRQ_PR);
  NVIC_EnableIRQ(AFE_EV_CNT_IRQ); 
  AFE_SPI->EVENTS_END = 0;
  AFE_EV_CNT->TASKS_START = 1;
}

void ZondFreqHwInit(void)
{ 
  F_SKIN_TIMER->MODE    = TIMER_MODE_MODE_Counter<<TIMER_MODE_MODE_Pos;
  F_SKIN_TIMER->BITMODE = TIMER_BITMODE_BITMODE_08Bit;
  F_SKIN_TIMER->CC[0]   = (1000/25)/2; //  25 Hz

  NRF_GPIOTE->CONFIG[F_SKIN_GPIOTE] |= GPIOTE_CONFIG_MODE_Task << GPIOTE_CONFIG_MODE_Pos;
  NRF_GPIOTE->CONFIG[F_SKIN_GPIOTE] |= GPIOTE_CONFIG_POLARITY_Toggle<<GPIOTE_CONFIG_POLARITY_Pos;
  NRF_GPIOTE->CONFIG[F_SKIN_GPIOTE] |= F_SKIN<<GPIOTE_CONFIG_PSEL_Pos;

  
  NRF_PPI->CH[F_SKIN_PPI_CH1].EEP	= (uint32_t) &NRF_GPIOTE->EVENTS_IN[AFE_DRDY_GPIOTE];
  NRF_PPI->CH[F_SKIN_PPI_CH1].TEP	= (uint32_t) &F_SKIN_TIMER->TASKS_COUNT;
  
  NRF_PPI->CH[F_SKIN_PPI_CH2].EEP 	= (uint32_t) &F_SKIN_TIMER->EVENTS_COMPARE[0];
  NRF_PPI->CH[F_SKIN_PPI_CH2].TEP 	= (uint32_t) &NRF_GPIOTE->TASKS_OUT[F_SKIN_GPIOTE];
  NRF_PPI->FORK[F_SKIN_PPI_CH2].TEP = (uint32_t) &F_SKIN_TIMER->TASKS_CLEAR;
 
  NRF_PPI->CHENSET = 1 << F_SKIN_PPI_CH1;
  NRF_PPI->CHENSET = 1 << F_SKIN_PPI_CH2; 
}

Function  AdsInitHW () is used at start to adjust the part.

The function AdsSetFrameConfig () is used after the program setting of the ADC to set the counting timer for the received data. On the ready data signal, the PPI lowers the CS signal and starts reading the data. Upon completion of the reading, the PPI raises the CS and increments the counter. When 100 is reached, an interrupt is called, in which the data is transferred to the handler (FreeRTOS stream) and switches the buffer to the DMA ArrayList.

The function ZondFreqHwInit()  is used to configure a rectangular waveform generator. By the event of readiness of data from the ADC, the counter is incremented. When the value 20 is reached, the output state is inverted and the timer is reset.

Everything works as intended, but ...
Logical analyzer recording showed a situation when the output inversion occurs not on the 20th event, but on the 19 (!). This behavior is repeated, but in a completely random way.

I attach the screenshots of the logical analyzer recording.

The screenshot shows that for some reason there was a reaction to 19 pulses. At the same time, the other systems work normally.

I can not understand with what it can be connected.

[Håkon Alseth]

Hi,

 

Is channel 10 counting based on the channel 8 signal (CSN of AFE_SPI)?

How random is this behavior? 1 in 10?

 

If one of the input signals is quite fast, you might be running into this errata:

IN event may occur more than once on input edge

 

Could you try adding this workaround to see if this has any effect on the behavior?

 

Kind regards,

Håkon

[CheMax]

Hi,

Yes, Channel 8 (AFE Data Ready)  is used as a base for signal generation. Signal parameters:
1. the duration at zero is approximately 12.5 microseconds
2. a period of 1 ms

The error occurs entirely by accident. It can occur 1 time in 10 minutes, it can 3 times in 2 minutes.

On the analyzer's record, there is no double front (sample rate was 16 MHz).

This can be the effect of impulse (> = 750 kHz) interference?

I'll try the solution you proposed and write about the results.

[CheMax]

So, I tested for an hour and a half of uninterrupted recording. This error did not appear.

changes except those described in the Erata *** I did not add.

Can you tell us what exactly this record:

*(volatile uint32_t *)(NRF_GPIOTE_BASE + 0x600 + (4 * GPIOTE_CH_USED)) = 1;

does?

The specification does not describe this region of memory. I understand correctly that this is an additional configuration register?

[Håkon Alseth]

Hi,

 

The specification does not describe this region of memory. I understand correctly that this is an additional configuration register?

This forces the peripheral clock to be always on in GPIOTE IN module, similar to how the GPIOTE IN behavior was on the nRF51-series devices.

 

This can be the effect of impulse (> = 750 kHz) interference?

It is very likely that this is the case. Did you see any discrepancies in the scoped signal that correlates with errata #155 ?

 

Best regards,

Håkon

[CheMax]

Hi,

It is very likely that this is the case. Did you see any discrepancies in the scoped signal that correlates with errata #155 ?

I Did not see. The analyzer does not show (checked at a sampling rate of 100 MHz), the oscilloscope is difficult to catch this moment.
Perhaps this is a consequence of the effects of static voltage.