SPI conflict with two shields - IOREF?

Question

Update:

I've refactored the code to be very resilient to boards missing or not being able to talk to them. Then I tinkered with a bunch of pins and learned that it all comes down to the IOREF pin being connected to the USB Host shield or not. If it's connected, the USB shield "takes over", allows communication with it, but the CAN shield does not work. If the USB shield IOREF pin is not connected, but the rest of the shield's pins are, CAN works. If I disable the IOREF pin going into the CAN shield (and thus the USB shield also since it's on top), the CAN shield works fine still, but the USB shield does not.

Something happens when the IOREF is powered into the USB board and if it's not powered, you can't talk to it.

Original post:

I have the following setup:

1. Yourduino RoboRED board
2. Seeed CAN Bus Shield v1.2 http://wiki.seeed.cc/CAN-BUS_Shield_V1.2/
3. USB Host Shield https://www.arduino.cc/en/Main/ArduinoUSBHostShield

My issue is that if I connect only the CAN bus shield, I can initialize it without any problem, but when I connect the USB host shield on top of the CAN, the CAN shield becomes unresponsive.

I looked into the wiring and per the specs, the following is how the hardware is set up:

• SPI bus runs on ICSP header for both shield, not on D11, D12 & D13. I could switch it for the CAN bus shield if I wanted/needed to, but I don't believe I shoudl be.
• The trigger pin for the USB host is D10
• The trigger pin for the CAN bus is D9 (since it's v1.2 of the board)
• Both boards have power. USB board is on top since it does not have an ICSP male connector on the top side, only female on bottom.

I started to debug the code and traced down to this CAN bus function in mcp_can.cpp, which I've instrumented with debug:

INT8U MCP_CAN::mcp2515_setCANCTRL_Mode(const INT8U newmode)
{
    Serial.print("mcp2515_setCANCTRL_Mode with newmode = ");
    Serial.println(newmode);
    INT8U i;

    mcp2515_modifyRegister(MCP_CANCTRL, MODE_MASK, newmode);

    i = mcp2515_readRegister(MCP_CANCTRL);
    i &= MODE_MASK;
    Serial.println("i & MODE_MASK");
    Serial.println(i);
    Serial.println(MODE_MASK);

    if ( i == newmode ) 
    {
        return MCP2515_OK;
    }

    return MCP2515_FAIL;
}

The output when it fails is:

USB started ok
past SPI begin
mcp2515_init called
mcp2515_setCANCTRL_Mode with newmode = 128
i & MODE_MASK
0
224
Enter setting mode fall
res = 1

For reference, this is the mcp2515_modifyRegister function:

void MCP_CAN::mcp2515_modifyRegister(const INT8U address, const INT8U mask, const INT8U data)
{
    #ifdef SPI_HAS_TRANSACTION
        SPI_BEGIN();
    #endif
    MCP2515_SELECT();
    spi_readwrite(MCP_BITMOD);
    spi_readwrite(address);
    spi_readwrite(mask);
    spi_readwrite(data);
    MCP2515_UNSELECT();
    #ifdef SPI_HAS_TRANSACTION
        SPI_END();
    #endif
}

The whole file is here: https://github.com/Seeed-Studio/CAN_BUS_Shield/blob/master/mcp_can.cpp. I've tried setting SPI_HAS_TRANSACTION to true with no change in the outcome, though I also don't know exactly what this does.

This is my init:

void setup (void)
{
    Serial.begin(115200);

    // init can bus : baudrate = 500k

    if (Usb.Init() == -1)
        Serial.println("OSC did not start.");
    else
        Serial.println("USB started ok");

    delay( 200 );

    while (CAN_OK != CAN.begin(CAN_500KBPS)) {
        Serial.println("CAN BUS Shield init fail");
        Serial.println(" Init CAN BUS Shield again");
        delay(10000);
    }

    Serial.println("CANbus ready");
}

It seems to have something to do with the registers available for write by the CAN bus library whenever the USB board is or isn't present. The write fails when the USB board is there. Any thoughts as to why or other things to try would be greatly appreciated.

Answer 1

There's a hardware conflict between these two boards which seems to not be resolvable. The solution was to buy the Keyes USB shield, make sure only it has access to pins 9 and 10 from the Arduino and then wire pin 7 (or 8) into pin 9 of the CAN shield, while also properly initializing the CAN shield with that new pin.

Lastly, the Keyes USB shield's ICSP header isn't long enough, so you need to cut 2 mm off the male non-ICSP pins in order to seat it properly on the CAN shield.