STM32F103 CAN Arbitration lost

[DNeum.1]

I have a problem using CAN on STM32F103C8T6. I use this MCU on a bluepill board,

a SN64HVD230 based CAN PHY and on the other end of the CAN bus a PiCAN2 board on

Raspberry Pi. I drew a picture to show my setup:

I want to have a 500 kbit/s-CAN-connection, I use the MCU with 16 MHz.

My initialization code (I use the STM32-HAL-API) for the CAN is the following:

#define CAN_RX_PORT GPIOA
#define CAN_RX_PIN LL_GPIO_PIN_11
#define CAN_TX_PORT GPIOA
#define CAN_TX_PIN LL_GPIO_PIN_12
 
void init_can_gpios(void){
 LL_GPIO_SetPinMode(CAN_RX_PORT, CAN_RX_PIN, LL_GPIO_MODE_INPUT);
 LL_GPIO_SetPinPull(CAN_RX_PORT,CAN_RX_PIN,LL_GPIO_PULL_UP);
 LL_GPIO_SetPinMode(CAN_TX_PORT, CAN_TX_PIN, LL_GPIO_MODE_ALTERNATE);
 LL_GPIO_SetPinSpeed(CAN_TX_PORT, CAN_TX_PIN, LL_GPIO_SPEED_FREQ_HIGH);
 LL_GPIO_SetPinOutputType(CAN_TX_PORT,CAN_TX_PIN,LL_GPIO_OUTPUT_OPENDRAIN);
}
 
void init_can_periph(void){
 LL_APB1_GRP1_EnableClock(LL_APB1_GRP1_PERIPH_CAN1);
 __HAL_RCC_CAN1_CLK_ENABLE();
 
 hcan1.Instance = CAN1;
 hcan1.Init.Prescaler=2;
 hcan1.Init.TimeSeg1=CAN_BS1_8TQ
 hcan1.Init.TimeSeg2=CAN_BS2_7TQ;
 hcan1.Init.Mode = CAN_MODE_NORMAL;
 hcan1.Init.SyncJumpWidth = CAN_SJW_1TQ;
 hcan1.Init.TimeTriggeredMode = DISABLE;
 hcan1.Init.AutoBusOff = ENABLE;
 hcan1.Init.AutoWakeUp = DISABLE;
 hcan1.Init.AutoRetransmission = ENABLE;
 hcan1.Init.ReceiveFifoLocked = DISABLE;
 hcan1.Init.TransmitFifoPriority = DISABLE;
 HAL_CAN_DeInit(&hcan1);
 HAL_StatusTypeDef retval = HAL_CAN_Init(&hcan1);
 HAL_CAN_Start(&hcan1);
 HAL_CAN_ActivateNotification(&hcan1, CAN_IT_RX_FIFO0_MSG_PENDING|CAN_IT_TX_MAILBOX_EMPTY);
 
}
 

The call of HAL_CAN_INIT(&hcan1) returns HAL_OK.

I try to send a message with:

void can_send_dummy_data(void){
 //We send on ID 123, 4 Bytes: 0x3456789A
 CAN_TxHeaderTypeDef tx_header;
 tx_header.DLC=4;
 tx_header.ExtId=0;
 tx_header.StdId=0x123;
 tx_header.IDE=CAN_ID_STD;
 tx_header.RTR=CAN_RTR_DATA;
 //Now data
 uint8_t tx_data[4]={0x34,0x56,0x78,0x9A};
 //Now send
 uint32_t tx_mailbox;
 HAL_StatusTypeDef retval=HAL_CAN_AddTxMessage(&hcan1,&tx_header,tx_data,&tx_mailbox);
}

The retval ist also HAL_OK. But no message appears on the other side of the CAN Bus.

When I look into the

CAN_MSR:

0b0000 0000 0000 0000 0000 1100 0000 0000

RXM and TXM are set

CAN-TSR:

0b0001 1100 0000 0000 0000 0000 0000 0100

Transmit Mailbox 2 empty

Transmit Mailbox 1 empty

Transmit Mailbox 0 empty

Arbitration lost to mailbox 0

CAN-ESR

0b1000 0010 0000 0000 0000 0000 00010011

Receive error counter = 130 -> Error passive

LEC = Stuff Error

Error Passive flag

Error warning flag

So I have arbitration lost and stuff error and error passive.

If I exchange my Bluepill assembly by an STM8-Discovery-Board running a CAN Hello

world application on STM8, I can send CAN-Messages (500 kbit/s) without problem.

So I assume I made mistakes in my CAN initialization code on STM32.

[DNeum.1]

Additional step: If I pull out the USB cable, which means depowering the bluepill board, I can see the CAN message on the bus. How to interpret this?

[Ozone]

Probably some potential shift relative to GND, because of different power supplies.

You could try to disconnect CAN_GND (leaving only CAN_HI and CAN_LO), which usually works as well.

[DNeum.1]

Finally, I got it. I power the bluepill board via an external 3V3 source and not via USB any more. It may have something to do with the double usage of pins PA11 and PA12 with USB. I tried to remap CAN-RX/CAN-TX it to the alternative PB8/PB9, but it did not work.

[Amel NASRI]

Hi @DNeum.1​ ,

I assume that the article "Concurrent use of USB and CAN with STM32F103" provides the explanation for what you see as strange behavior.

-Amel

[DNeum.1]

Since I used only powering by USB, but not the USB peripheral, I thought it would be no problem. But it was... Thank you for your patience.