Safety: Upcoming change in Watchdog

[Peter-van-Tol]

Discussed in #62

^{Originally posted by Peter-van-Tol December 10, 2023}
There are some upcoming changes in the watchdog, for safety reasons:

• bug resolved in the driver. The signal when the watchdog has bitten was not relayed to the HAL.
• if the watchdog has bit and the has_bitten bit is True, the user can reset it to False to resume operation.
• added an physical enable_out pin to the watchdog. Which pin it has to be, can be chosen in the configuration (see example below). I use this pin with the HUB75HAT project. When the board is running enable_out is HIGH. When the watchdog times out (either due to communication disruption or LinuxCNC stops sending signals) the pin becomes LOW. On the HUB75HAT this means that all input and output will go in high-Z mode. This means the watchdog safety is completely in hardware and does not depend on LinuxCNC or the communication.

Example configuration (default for HUB75HAT):

{
    "board_name": "EMCO5_SPI",
    "board_type": "HUB75HAT v8.0",
    "clock_frequency": 50000000,
    "connection": {
        "connection_type": "spi",
            "mosi": "spi:0",
            "miso": "spi:1",
            "clk":  "spi:2",
            "cs_n": "spi:3"
    },
    "watchdog": {
        "pin":"ena:0"
    },
    "modules": [
        ...
    ]
}

Two open actions before I commit this change are:

• making the physical enable_out pin optional;
• create a working example where the watchdog is part of the E-stop latch circuit. This requires some additional components, to drive the has_bitten pin low before enabling the E-stop chain.

[Peter-van-Tol]

This issue only addresses the watchdog setting the external ENA-pin. External triggers (i.e. E-stop completely within the FPGA) will be another issue.