CDBUS Bridge

The CDBUS Bridge HS is a USB-to-RS485 (CDBUS) adapter. The USB port supports High-Speed, and the RS485 interface supports up to 50 Mbps.

• The two RS485 ports are internally straight-through, simplifying wiring.

Switchs Defination:

• S1.1: Force bootloader mode.
• S1.2: In arbitration mode, the maximum limit of baud_l:
  OFF: Default 1 Mbps; ON: Default 2 Mbps. (modifiable)
• S2.1: Enable pull-up resistor.
• S2.2: Enable termination resistor.
• S2.3: Enable pull-down resistor.
• S2.4 (HW v6.2+): When enabled together with S2.3, pulls B to 1.65V, allowing signal A to operate as a TTL single-wire serial.
• S2.5 (HW v6.2+) / S2.4: Enable 5V output (should disable when using external power supply).

Two Ports, One Firmware

The bridge presents both a serial port and an ethernet port, always, and either may be used. Nothing has to be reflashed to switch between them.

One of them has the bus at a time: the serial port while it is open, the ethernet port otherwise. Sending works from either. Opening the serial port is something you do on purpose, while the ethernet interface tends to come up on its own, so the serial port wins and the existing tools keep working without anything being taken down first.

Serial Port (unchanged)

• The PC sends complete CDBUS packets (with CRC) via USB serial to the CDBUS Bridge, which forwards them unchanged to the RS-485 bus.
• Data received from RS-485 is sent unchanged back to the PC via USB serial.
• The baud rate set by the PC when opening the USB serial port is used for RS-485 (baud_l is automatically limited in arbitration mode).
• Before that, the bus runs at the rate stored in the config, which is what it comes up with after power on. Saving the config stores the rates in effect at that moment, so a save made while the port is open at 2 Mbps makes 2 Mbps the boot rate.
• The PC must enable the DTR option on the USB serial port. The port counts as open for as long as DTR is asserted, and DTR is deliberately not cleared by a USB reset: if the host resets the bus while a program holds the port (a wake up, a hub replug), the bridge keeps treating the port as open, and the ethernet port does not get the bus back, until the program closes and reopens it.
• The default RS-485 address of the Bridge is 0. To change it, see below.
• Raw mode allows arbitrary data transfer without following the CDBUS byte format (HW v6.2+). The bus then belongs to the serial port alone and the ethernet port reports no carrier.

Ethernet Port

The whole bus is mapped into fdcd::/104: the last 3 bytes of an IPv6 address are the CDNET address level:net:mac and the UDP port is the CDNET port. Talking to a device is therefore plain IPv6 UDP.

• Linux, macOS and Windows 11 have an in-box driver (CDC NCM), nothing to install.
• No daemon on the host, and no port to open exclusively: several programs can use the bus at the same time.
• Bus traffic goes here whenever the serial port is not open. Close it, or leave it alone, and the ethernet port has the bus.
• A host that is not reading the port (interface down, driver not bound) is noticed within 200 ms; bus traffic is then dropped rather than queued for it, and queuing resumes the moment the host reads again.
• The bus comes up at the rate stored in the config, so the serial port does not have to be opened once just to pick a rate for this one. To change it, write bus_cfg_baud_h, save and power cycle.
• The address the host holds is the bridge's identity on the bus, so it has to match bus_cfg_mac and net on the device (00 and 00 by default).
• A CDNET packet has to fit in one CDBUS frame and there is no fragmentation, so keep datagrams at 244 bytes or less.
• port_offset shifts the host's own port: a program binds port_offset + <CDNET port> and is that CDNET port on the bus. It is 0xcd00 (52480) by default, so a level 0 port, which is under 128, needs no root to bind; 0 switches the shift off. A datagram sent from a port below the offset is dropped, so bind explicitly rather than rely on an ephemeral port.

It needs a one-time host setup, see fw_bridge/host/. Until that is done the serial port works as it always has, so the ethernet port is opt-in.

s = socket.socket(socket.AF_INET6, socket.SOCK_DGRAM)
s.bind(("fdcd::80:0", 0xcd00 + 40))     # port_offset + CDNET port 40
s.sendto(b"...", ("fdcd::80:00fe", 0xcdcd))

Configuration

CDBUS GUI Tool: https://github.com/dukelec/cdbus_gui

The same services (port 1: device info, port 5: config status area, port 8: flash) are reachable two ways:

• Serial: open the port with the baud rate 52685 (0xcdcd) and set the target address to 00:00:ff. The bus is not touched while the port is in this mode.
• Ethernet: fdcd::10:0, the one address inside the prefix that never reaches the bus. It is a node the firmware serves itself, and it answers to whichever of the host's addresses the request came from.

After modifying the configuration, write 1 to save_conf to save the changes to flash.

To restore the default configuration, change the value of magic_code to a different value, save it to flash, and then power cycle the device.

Debug output always goes out on the debug uart, and dbg_en sends a copy to the host on CDNET port 9: bit 0 (1) to the serial port, bit 1 (2) to the ethernet port, from fdcd::10:0 to the host's level 0 address, 3 for both. The serial port is there from the moment the device enumerates, so the boot log, the CSA table included, waits in its queue and is still there when the port is opened; if the port is never opened, bus traffic takes those frames back as it needs them. The ethernet port has no open to wait for, so it only carries what is printed while it is up and being read.

The debug uart (2 Mbps) is fed through a 2 KB ring buffer drained by DMA, so printing never holds up the main loop. The ring survives a reset: whatever a crash or a watchdog left unsent is printed by the next boot, between --- unsent before reset --- and --- end ---, followed by the reset cause. A hard fault prints the fault registers and the faulting pc before it hangs, and a deliberate reset flushes the ring first.

Download Source Code

git clone --recursive https://github.com/dukelec/cdbus_bridge

For other hardware versions, please switch to the corresponding branch.

The tinyusb submodule is pinned at an upstream commit rather than at a release tag, because 0.21.0 predates a batch of usbd.c fixes for ways the device can wedge until it is replugged. Nothing extra to run.

Test

git clone --recursive https://github.com/dukelec/cdbus_tools

cd cdbus_tools/
./cdbus_terminal.py --help

We can also use a generic serial debugging tool for loopback testing. For example, to simulate node 02 sending a packet to node 00 (with CRC):

02 00 01  cd  c1 99

If the hardware works properly, the same packet should be received.