Custom MCP2515+TJA1042T board stuck on sending

[digitalEnthusiast]

Hi,

I am working on a project that requires an easy to implement yet robust communication protocol. I chose CAN because it seemed to fit. The data that must be transmitted is fairly simple. It has a node identifying byte and a data byte. In a prototype I used one the popular dev boards for CAN bus shown below and it worked and is still working without major issues. I has an MCP2515+TJA1050.

Once everything was working well I decided to integrate the communication hardware into the project PCB. I will not get into details here but it is a type of keyboard, so no high power components, no RF components.
For reference I used the module schematic as well as datasheets and design notes for MCP2515 and the TJA1042T (the one I was able to get). The communication part of the circuit is this:

The strange looking component 2WTS1 (2 way toggle switch) is a sliding switch with two sets of contacts to be able to choose between the two 60ohm resistors and the two 1.3k (or nothing). It looks like this:

The oscillator is an integrated one like this:

The test code that I used to see if everything works is the following one:

#include <Arduino.h>
#include <SPI.h>
#include <Wire.h>
#include "CAN.h"
#include "PCF8574.h"

void send(uint8_t item, uint8_t value) // as name indicantes, simple funcion to send the data structure
{

  uint8_t id;
  const uint8_t id_sender = 0xA0; // left KB

  id = id_sender + item;
  Serial.println("Data ready00");

  CAN.beginPacket(id, 1);
  Serial.println("Data ready01");
  CAN.write(value);
  Serial.println("Data ready02");
  CAN.endPacket();
  Serial.println("Data ready03");

  Serial.println("Data sent");
}

void setup()
{

  //delay(2000);
  Serial.begin(9600);
  Serial.println("Puto el que lee");
  int i;

  pinMode(9, OUTPUT);

  digitalWrite(9, HIGH);

  
  Wire.begin();
  //i2c_scan();

  CAN.setPins(10, -1);
  CAN.setClockFrequency(8E6);
  
  
  if (CAN.begin(500E3))
  {
    Serial.println("CAN started!!!");
  }
  else
  {
    Serial.println("CAN failed!!!");
  }

}

void loop()
{

  int packetSize = CAN.parsePacket();
  
  send(2, 0xA0);
  delay(500);

  ////------- RECEIVER -------////

  
    if (packetSize)
    {
      // received a packet
      Serial.print("Received ");
      Serial.print("packet with id 0x");
      Serial.println(CAN.packetId(), HEX);

      while (CAN.available())
      {
        received = CAN.read();
        Serial.print(received, BIN);
        Serial.print(" - ");
        Serial.print(CAN.packetId(), HEX);
        Serial.println();

      }
    }
    
}

There are too many Serial prints, but i was trying to find the failure point.
So, the sender always freezes right after "Serial.println("Data ready02");", meaning the issue is the "CAN.endPacket();". I was not able to make the custom boards from the schematic above move past it. The dev board from the first image works without issue.
I did a little digging in the library itself and found that the issue is in the "while" loop on line 191 of MCP2515.cpp:

while (readRegister(REG_TXBnCTRL(n)) & 0x08) {
    if (readRegister(REG_TXBnCTRL(n)) & 0x10) {
      // abort
      aborted = true;
      
      modifyRegister(REG_CANCTRL, 0x10, 0x10);
      
    }
    //Serial.println(readRegister(REG_TXBnCTRL(n)), BIN);
    //Serial.println(aborted);
    //Serial.println("Data ready02a");
    yield();
  }

The result of the "Serial.println(readRegister(REG_TXBnCTRL(n)), BIN);" was 101000 and once 1101000. According to the MCP datasheet, this corresponds to Message transmit Request (logical), Message Lost Arbitration and Message Aborted Flag.
This is as far as I could get. I do not have an osciloscope, so I cannot look at the data between the ICs.

The following was tried without success:

• taking out the switch for changing between resistor sets
• adding a 10k resistor between the STB pin of TJA and ground
• checked continuity multiple times, as well as the welds and presence of shorts

Does anyone have any idea of what might be the issue here? I still have the possibility of using the dev board for communication. It looks bad and needs a lot of space compared to the on-board solution but at least it works. I really want to be able to integrate CAN bus hardware into my own circuits, but this problem is driving me crazy.

Cheers