Register Definitions

A board’s register layout lives in config/<project>.yaml, where <project> matches the firmware directory name under src/.

Schema

Key	Required	Meaning
struct_name	yes	name of the generated C++ struct, e.g. abs_config_t
chip	yes	part number, used for the flash geometry.
regs	yes	ordered list of registers. Order determines addresses
can_filtering	yes	FDCAN hardware filter and timing settings

regs

Key	Required	Meaning
name	yes	register name, uppercased for the C++ member
type	yes	see the type table below
fields	no	bit fields within the register

Each entry under fields has a name and a pos (bit position, counting from the LSB). A register with fields is addressed only through those fields; one without gets a whole-register accessor instead.

can_filtering

Key	Required	Meaning
id_reg	yes	register holding this node’s CAN ID. Must name a declared register
id_type	yes	ext is the only supported value
delay_compensation	yes	transceiver delay compensation, needed for bit rate switching
tdc_offset	yes	transmitter delay compensation offset
tdc_filter	yes	transmitter delay compensation filter window
can_subs	no	additional subscribe filters, each with can_id and id_type

Types

YAML type	C++ type	Bytes	Generates header	Can be flashed
uint8	uint8_t	1	yes	yes
uint16	uint16_t	2	yes	yes
uint32	uint32_t	4	yes	yes
float32	float	4	yes	yes

Size Limit

The whole config must fit in one flash page. On the STM32G431CB that is 2048 bytes, and the generator raises an error if the registers exceed it.

Example

config/abs.yaml:

struct_name: abs_config_t
chip: STM32G431CBTx
regs:
- name: can_id
  type: uint8
- name: host_can_id
  type: uint8
- name: sys_cfg
  type: uint16
  fields:
  - name: continuous_mode
    pos: 0
  - name: bounded_mode
    pos: 1
  - name: invert
    pos: 2
- name: output_scalar
  type: float32
can_filtering:
  id_reg: can_id
  id_type: ext
  delay_compensation: true
  tdc_offset: 13
  tdc_filter: 1

Addresses fall out of the order: can_id at 0x0, host_can_id at 0x1, sys_cfg at 0x2 (two bytes), output_scalar at 0x4.

Generated Header

config_gen.py renders this into src/abs/Inc/abs_config.hpp, which defines:

struct abs_config_t {
    reg_t<uint8_t>  CAN_ID{0x0};
    reg_t<uint8_t>  HOST_CAN_ID{0x1};
    reg_t<uint16_t> SYS_CFG{0x2};
    reg_t<float>    OUTPUT_SCALAR{0x4};

    using continuous_mode = field_t<&abs_config_t::SYS_CFG, 0>;
    using bounded_mode    = field_t<&abs_config_t::SYS_CFG, 1>;
    using invert          = field_t<&abs_config_t::SYS_CFG, 2>;
    using output_scalar   = field_t<&abs_config_t::OUTPUT_SCALAR>;

    template<typename F> auto get() const;
    template<typename F> void set(auto value);

    auto set_raw(uint8_t address, uint32_t raw) -> bool;
    auto get_raw(uint8_t address, uint32_t& raw) const -> bool;
};

Firmware reads and writes it by field:

if (config.get<abs_config_t::invert>()) {
    position = -position;
}

set_raw and get_raw dispatch by address and are what the CAN handler calls. Values persist in the last flash page of the chip: a write reads the page into RAM, patches the register, erases the page and reprograms it.

A compile-time validator checks that no two registers overlap, so a malformed layout fails the build rather than corrupting flash.

Adding a Register

1. Append it to regs in config/<board>.yaml. Append, do not insert, since inserting shifts every later address and invalidates every already-configured board.
2. Rebuild the firmware so the header regenerates.
3. Add the key to each device file under rover/ that needs a non-default value.
4. Reflash and reconfigure the affected boards.