Possible to read but can't send data.

[Toernblom]

Hi, currently using STM32F103R8T6.

CAN interface is connect like:
PB8 -> CAN_RX -> TJA1050
PB9 -> CAN_TX -> TJA1050

in void Setup I'm using:
bool ret = CANInit(CAN_500KBPS, 2); // CAN_RX mapped to PB8, CAN_TX mapped to PB9

What happens is that while debugging i notice that on this line, the while loop never ends.

Communication works ok outside this unit. Am i missing something?

[Toernblom]

Errors while running. I've used the same cables with another board using CAN so shouldn't be an issue. Terminations is ok aswell because other devices can talk to each other.

[nopnop2002]

Regarding send fail, past issues (closed issues) are helpful.

Mostly due to terminating resistors or cable quality.

Many people have encountered this problem and solved it by changing the terminator or STAB wire cable.
When STM32 is added to CAN-BUS in operation, the impedance may change due to STAB.
When adding nodes, pay attention to the length of the STAB cable.

[Toernblom]

I currently have 3 nodes, 2 of them are talking and all has twisted shielded pairs which are 5-10cm in length and the shield is grounded. It's the only device on the network that can't talk. The resistance between CAN_L & CAN_H is ~60 Ohm so that isn't an issue either. I've also checked previous closed issues and tried those aswell. All baudrates is 100kbps to minimize noise.

These are the prints from your code. I noticed that the peripheral_clock_rate was at 32mhz while STM32F103R8 that i'm using is at 72mhz so i tried changing that aswell to the correct clock with no success

[nopnop2002]

peripheral_clock_rate is not cpu frreuency.
cpu frreuency = peripheral_clock_rate x 2

Arduino core support for STM32 based boards uses different clock frequencies depending on the Board Part Number.
BlackPill & BluePill use 72Mhz.
Generic board use 64Mhz.

If the receive works correctly, it's not a CPU frequency issue.

The resistance between CAN_L & CAN_H is ~60 Ohm

I think It's too small.

[Toernblom]

Hmm okey, then peripheral_clock_rate should be 36mhz then? Since STM32F103R8T6 is 72Mhz not 64.

The receive did work yesterday, not anymore for some reason.. I saw your previous comment on another issue with 150 Ohm termination, I'll try that!

[nopnop2002]

Since STM32F103R8T6 is 72Mhz not 64.

peripheral_clock_rate is not cpu frreuency.
cpu frreuency = peripheral_clock_rate x 2

Arduino core support for STM32 based boards uses different clock frequencies depending on the Board Part Number.
BlackPill & BluePill use 72Mhz.
Generic board use 64Mhz.

You can check your board CPU freqiency using the F_CPU macro

Serial.println(F_CPU);

[Toernblom]

Hmm, that result gave 64 Mhz. In my Platformio.ini file , the board is genericSTM32F103R8.
Should i switch out the board part number?

[nopnop2002]

Should i switch out the board part number?

If you use bluepill or blackpill in Platformio.ini, it will be 72Mhz.

It has nothing to do with this issue.

But everyone gets confused.

[Toernblom]

Haha i understand why. We use a custom made PCB with STM32F103R8T6 and It's the first time with the STM32 chip family. So I thought i'd should use the genericSTM32F103R8 in the .ini file , not blackpill or bluepill ..

But if i use the genericSTM32F103R8 F_CPU shows 64Mhz

[nopnop2002]

I think you can build custom boards using BluePill and BlackPill variants.

However, the GPIO of the Serial object changes for the Generic board, BluePill, and BlackPill.

The output destination of Serial.print differs depending on the Board Part Number.

BluePill:PA9
BlackPill:PA9
Generic F103:PA2

everyone gets confused.

[Toernblom]

Can't use bluepill or blackpill in platformio.ini file to make serial work. If i use genericSTM32F103R8 with HardwareSerial its fine.

its because

However, if i do use genericSTM32F103R8 F_CPU is wrong and if i use either bluepill or blackpill in .ini the f_CPU is correct..

I do understand that everyone gets confused..

[Toernblom]

I have replaced the termination resistors with 150 Ohm so that i have two terminations at the ends with a bus resistance at 75 Ohm now. No success in sending any data.

[nopnop2002]

Keep STAB length as short as possible.

[Toernblom]

This is the setup, the connection between the raspberry pi on the right and the CAN analyze tool on the bottom has no shielded nor twisted pairs. I'm not sure the cables are the issues..

[nopnop2002]

If you have a BluePill or BlackPill, you can identify the problem.

Try BluePill or BlackPill.

[Toernblom]

Sadly I don't. I can purchase one however if you think that would help ?

[nopnop2002]

There may be a problem with your custom made PCB.

If it works fine with BluePill/BlackPill, then there may be a problem with your custom made PCB.

BluePill/BlackPill works absolutely.

[Toernblom]

The connection:

Your code says that the initialization was ok. I don't think theres anything wrong with the wiring here..

[Toernblom]

Alright, found out the reason why reading stop worked. It was cause i was monitoring on the UART. If i only powered it on and ran with debug i could read messages. Not send however. Its quite weird actually.

If dont use Hardware Serial at PA10 & PA9

Then the code gets stuck here
CANInit(CAN_250KBPS, 2);

Stepping over CANInit function with debugging results in the debug can't step over and gets stuck inside this function
ComputeCANTimings(HAL_RCC_GetPCLK1Freq(), target_bitrate, &timings);

Digging further it goes to stm32f1xx_hal_rcc.c
and gets stuck here

uint32_t HAL_RCC_GetPCLK1Freq(void)
{
  /* Get HCLK source and Compute PCLK1 frequency ---------------------------*/
  return (HAL_RCC_GetHCLKFreq() >> APBPrescTable[(RCC->CFGR & RCC_CFGR_PPRE1) >> RCC_CFGR_PPRE1_Pos]);
}

And if i do use Hardware Serial at PA10 & PA9
the code doesn't hang on that line and continues running and i can successfully read messages on the CAN bus. But CANSend hangs here:

while (CAN1->sTxMailBox[0].TIR & 0x1UL && count++ < 1000000);

[nopnop2002]

It was cause i was monitoring on the UART.

This is the UART output.
Something is wrong.

Testing on a custom PCB is even more confusing.
Testing on a custom PCB doesn't reveal anything.

[Toernblom]

Alright, If i put a delay between Serial.begin and CANInit the code doesn't hang anymore. Still getting Send Fail no matter what i do, I've got another CAN Node with the same tranciever, TJA1050 , exact same cables and length on the network that is working well, custom made PCB aswell..

[amitesh-singh]

@Toernblom have you found the solution?

[nopnop2002]

@Toernblom
@amitesh-singh

It's not a clone chip, right?

Doesn't work properly with clone chips.

https://hackaday.com/2020/10/22/stm32-clones-the-good-the-bad-and-the-ugly/

[Toernblom]

Hi, it was a fault within the PCB. I think it was some issue with different logic levels between the IC's of the MCU and CAN controller & transciever.

[amitesh-singh]

@Toernblom @nopnop2002 Thanks for the reply.

Is it due to powering tja1050 by 5v and stm32f103 is powered from 3.3v?

[amitesh-singh]

cc @Toernblom @nopnop2002

[amitesh-singh]

the PB8 and PB9 pins are 5v tolerant and tja1050 should work with 3.3v devices, that's what they claim in their datasheet as tja1050 RXD is a totem pole and voltage varies from 0.7Vcc to 0.45v

[nopnop2002]

@amitesh-singh

the PB8 and PB9 pins are 5v tolerant

Yes. You are right.

I powered the STM32F103 from USB and tried it with the 5V transceiver TJA1051.
The terminating resistor worked correctly with both 100 ohm and 150 ohm.
My TJA1051 has components marked 4017V installed.
I don't know what this is.

My test circuit.
The Nano on the left is a receiver with a terminating resistor.
The Nano on the right is a transmitter without a terminating resistor.
TJA1051 has a terminating resistor.

This is the receiver logging

When CAN-BUS has a mix of 3.3V and 5V transceivers, +1V Ground Shift with Split Termination is effective
https://www.ti.com/lit/an/slla337/slla337.pdf